IAI Researchers Have 33 Papers Accepted at NeurIPS 2026

@conference{nokey,

title = {A3Bench: A Benchmark for Compositional Reasoning over Aggressive Interactions in Videos},

author = {Austin C Baker and Raiyaan Abdullah and Daniel Z Yoffe and Shruti Vyas and Yogesh Rawat

},

url = {https://openreview.net/forum?id=iWKda5S9fT

https://neurips.cc/virtual/2026/loc/atlanta/poster/139138},

year  = {2026},

date = {2026-12-01},

urldate = {2026-12-01},

publisher = {NeurIPS 2026},

abstract = {Recent Multimodal Large Language Models (MLLMs) have achieved strong performance on general video understanding benchmarks, yet their ability to reason about participant roles in aggressive interactions remains largely unexplored. Existing violence and crime datasets primarily evaluate whether aggression occurs, without testing whether models can identify who performed the action, who received it, and how participants are related within the scene. To address this gap, we introduce **A

Bench**, a benchmark to study *aggressive action and association* in videos. The benchmark evaluates whether models can jointly recognize aggressive actions, distinguish aggressors from victims and bystanders, and correctly bind actions to participants under challenging distractor settings. We benchmark 15 recent MLLMs and find that current models struggle substantially on fine-grained aggressive-scene understanding. Performance degrades consistently as reasoning complexity increases, with models frequently confusing aggressors with victims or bystanders, revealing a systematic failure in role-action association rather than coarse aggression recognition alone. We further observe that increasing model scale or applying generic reasoning prompts provides limited improvement. Finally, we explore a simple role-graph prompting strategy that encourages structured reasoning over participant interactions, leading to consistent gains on compositional reasoning. Our findings highlight that reliable aggressive-scene understanding remains a major challenge for current MLLMs and motivate future research on role-aware video reasoning beyond coarse violence detection.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{nokey,

title = {MaRiO: Multi-agent Collaborative Reasoning via Shared Observations in MLLMs},

author = {Nathaniel Redmond and Fidel Omar Tito Cruz and Devansh Sharma and Shehreen Azad and Sirshapan Mitra and Shruti Vyas and Yogesh Rawat

},

url = {https://neurips.cc/virtual/2026/loc/atlanta/poster/139083},

year  = {2026},

date = {2026-12-01},

urldate = {2026-12-01},

publisher = {NeurIPS 2026},

abstract = {Multimodal Large Language Models (MLLMs) show strong visual perception capabilities, but existing evaluations largely focus on single-view or multi-view settings with shared camera parameters, leaving the challenge of {multi-agent collaborative} scenarios underexplored. In such settings, multiple agents observe a shared environment from independent viewpoints, and reliable decisions require reasoning across these independent observations of the same environment. To study this problem, we introduce MaRiO, a benchmark for evaluating multi-agent collaborative reasoning, and evaluate 23 state-of-the-art MLLMs on it. Our analysis reveals three key findings: (1) spatial tasks are generally easier than geometric reasoning tasks; (2) most models struggle to resolve cross-agent correspondences, with errors increasing with scene complexity; and (3) models show limited ability to leverage geometric cues present in the scene. As an initial step toward addressing these challenges, we explore geometry-guided synthetic view augmentation that generates intermediate transition frames between agent observations, providing auxiliary spatial context that improves performance in certain categories. Overall, our results highlight that multi-agent collaborative reasoning remains an open challenge and an important direction for multimodal embodied systems.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{nokey,

title = {PHOEBI: An Open-World Benchmark for Bacterial Identification in Phase-Contrast Microscopy},

author = {Aaditya Baranwal and MD JAHID HASAN and Shruti Vyas},

url = {https://arxiv.org/abs/2606.22890},

year  = {2026},

date = {2026-12-01},

publisher = {NeurIPS 2026},

abstract = {Optical microscopy lets us watch living bacteria in real time without stains or labels, and it remains the standard instrument for identifying species across clinical, environmental and industrial microbiology. Yet real samples rarely hold a single species. They arrive as mixtures and sometimes carry an organism no model has ever seen. What happens when a model trained on familiar mixtures meets one it has never encountered? No benchmark has asked this of phase-contrast images, so we built one. textbf{PHOEBI} brings together 120,000 phase-contrast images of 40 mixtures of six bacterial species that we cultured and imaged ourselves, and it holds whole mixtures out of training so that every test is a combination the model has never seen. The answer is sobering. Classifiers that are nearly perfect on the mixtures they know collapse on new ones, and no choice of backbone or pretraining rescues them. The features turn out to be sound and the failure lies in how the evidence within each image is pooled. A culture looks the same everywhere, and decoders built on this simple fact hold steady where large networks fail even though one of them is never trained by gradient descent. The same features go further still. They notice when a species outside the catalogue appears, and the simplest way of giving it a class of its own proves the most effective while leaving the known species almost untouched. PHOEBI turns the mixed culture, the everyday case at the bench, into a measurable target for open-world recognition.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{nokey,

title = {Re:Cognize: A Framework for Open-Set Sequential Character Re-Identification},

author = {Aaditya Baranwal and Madhav Kataria and Yogesh Rawat and Shruti Vyas},

url = {https://arxiv.org/abs/2609.34032},

year  = {2026},

date = {2026-12-01},

publisher = {NeurIPS 2026},

abstract = {Comics and manga tell stories through characters whose appearances change from panel to panel, yet readers recognise them hundreds of pages later without ever consulting a cast list. Character re-identification models rarely face the same challenge: existing evaluations give them a fixed set of identities before the story begins. textbf{Re:Cognize} instead asks models to recognise characters as the story unfolds and learn who is who along the way. Across three manga corpora, we evaluate five Re-ID backbones and textbf{MECHA}, a memory-augmented baseline we introduce, and find that the central difficulty lies somewhere unexpected. Recognising familiar characters is nearly solved. Knowing which matches to trust is not. When models build a gallery and grow it from their own predictions, their performance deteriorates, even though correct identity labels would improve it by over twenty points. This raises a simple question: when should a model trust its own match? We derive a single comparison that determines when gallery updates improve recognition and use it to design textbf{VOUCH}, which updates its memory only when the page provides supporting evidence. Without fitting additional parameters, VOUCH recovers between one-third and two-thirds of the improvement available from perfect labels. RE:COGNIZE measures whether a model can recognise characters as it reads; VOUCH shows how to remember them reliably. We focus our methodological claims on identity maintenance for previously encountered characters, while separately evaluating new-character emergence as a diagnostic under a fixed reference rule to distinguish recognition from discovery.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{mcgee2026trust,

title = {Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning},

author = {Cristian McGee and El Houcine Bergou and Aritra Dutta},

url = {https://openreview.net/forum?id=ZKbACxi4Qb},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Step-size selection remains a central challenge in large-scale neural network optimization: conservative steps slow convergence, while aggressive steps can destabilize training. We propose Zero-and-First-Order optimization (ZFO), a lightweight framework that decouples direction selection from step-size selection. ZFO uses a trusted first-order optimizer to determine the direction and performs zeroth-order evaluations only along this one-dimensional subspace to choose how far to move. Using the current forward-backward pass and two additional fixed-batch function evaluations, ZFO constructs a local model of the objective along the proposed direction and selects a curvature-aware step within a bounded search interval. This yields an adaptive step-selection mechanism with lower cost than full line search. We provide theoretical guarantees showing that shared-sample evaluations produce reliable finite-difference curvature estimates, that the induced local model selects a near-optimal step along the search interval, and that ZFO converges to a neighborhood of a stationary point. Across multiple language models and datasets, ZFO improves accuracy and stability over standard first-order fine-tuning algorithms.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{li2026attend,

title = {Attend Locally, Remember Linearly: Linear Attention as Cross-Frame Memory for Autoregressive Video Diffusion},

author = {Kunyang Li and Mubarak Shah and Yuzhang Shang},

url = {https://openreview.net/forum?id=BScHl5MZ8h},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Autoregressive (AR) video diffusion is a powerful paradigm for streaming and interactive video generation. However, its reliance on softmax self-attention leads to quadratic compute complexity in sequence length and memory usage due to key-value caching, which limits its scalability to long video horizons. Existing remedies (e.g., sparse attention and KV-cache compression) reduce per-step cost but still rely on a linearly growing cache or irreversibly discard past context, and thus fail to address linear memory growth and streaming context management. To address this scalability bottleneck, we propose ARL2 (Attend Locally, Remember Linearly), a hybrid attention module that replaces quadratic cross-frame attention with a fixed-size recurrent state. We decompose self-attention into two branches: an intra-frame softmax branch for spatial detail and local dependencies, and an inter-frame gated recurrent linear branch that maintains a fixed-size state for streaming context. Our key insight is that softmax attention captures fine-grained local interactions, while a recurrent state provides controllable long-range memory. This design achieves linear-time scaling with constant memory while improving temporal consistency over the full-softmax model. To prevent noisy intermediate states from corrupting memory, we update the recurrent state only after the denoised pass. To avoid within-frame information asymmetry, all tokens share the same pre-update state rather than sequential updates. To the best of our knowledge, this is the first work to convert a pretrained AR video diffusion model into a hybrid linear attention architecture, through an efficient two-stage training scheme for AR video. With 75% of layers replaced by hybrid linear attention, the model achieves up to 2.26 wall-clock speedup and 54% memory reduction, while maintaining comparable quality with improving temporal consistency.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{sun2026dynamically,

title = {Dynamically Structured Diffusion Language Model Decoding via Bayesian Inference},

author = {Bian Sun and Kevin Zhai and Mubarak Shah and Zhenyi Wang},

url = {https://openreview.net/forum?id=Z6jLmoD11G

https://arxiv.org/abs/2605.09820},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Diffusion language models (DLMs) have recently emerged as a promising alternative to autoregressive models, primarily due to their ability to enable parallel decoding. Despite this advantage, most existing DLMs rely on a fixed generation length specified prior to decoding, which restricts their flexibility in real-world applications. While a few recent works attempt to support flexible-length generation, they typically suffer from notable limitations: some require costly retraining to accommodate variable-length outputs, while others depend solely on local confidence signals during decoding. Such local criteria fail to capture the evolving structure of the sequence, often resulting in suboptimal generation quality. In this paper, we propose a training-free, Bayesian structured decoding framework that formulates flexible-length generation as a dynamic structural inference problem. Our approach formulates flexible-length generation as a dynamic structural inference problem, jointly computing the expansion length, the block boundaries, and the decoding schedule. At each window expansion step, the method integrates local uncertainty with structural signals via a unified mechanism that supports dynamic structured generation, including both flexible block expansion and block organization, while maintaining coherence. Extensive experiments across multiple benchmarks demonstrate that our approach significantly improves generation quality and flexibility over existing fixed-length and flexible-length baselines. These results highlight the advantage of Bayesian structured decoding for diffusion language model, providing a principled and efficient solution for structured text generation.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{siddiqui2026flashssm,

title = {FlashSSM-3D: A Distilled State Space Model for Lightning-Fast Dense 3D Reconstruction},

author = {Nyle Siddiqui and Mubarak Shah},

url = {https://openreview.net/forum?id=qrND6erSwi

https://nylesiddiqui.github.io/FlashSSM-3D-prefinal.pdf},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {3D geometric understanding is undergoing a major transition where rigid, deterministic, optimization-based pipelines are being replaced by large-scale, deep learning models. While recent methods achieve impressive performance on tasks such as multi-view 3D reconstruction and camera pose estimation, they suffer from two critical limitations: (i): current state-of-the-art (SOTA) models require industrylevel compute and training corpora, severely limiting accessibility, and (ii): recent approaches remain constrained to modifying existing pretrained transformer models, since substantially changing their architecture would require retraining from scratch. We instead study a complementary direction: training an attention-free state space model for dense 3D understanding. We introduce FlashSSM-3D, the first 3D reconstruction state space model capable of performing accurate geometric reasoning over hundreds of images with linear computational complexity. However, training such a model from scratch with competitive performance is impractical under typical academic compute constraints. To address this, we train FlashSSM-3D by distilling knowledge from VGGT, avoiding the need to learn 3D geometric priors entirely from scratch. However, standard distillation losses are poorly suited for dense 3D reconstruction, where preserving real-world spatial structure is essential. We therefore propose a geometry-aware knowledge distillation method specifically tailored for large-scale 3D reconstruction models. Our method introduces two novel losses: a Procrustes-based alignment loss that enforces global structural consistency between student and teacher 3D point predictions, and a camera reprojection loss that directly supervises view-consistent camera geometry between teacher and student. Extensive experiments across multi-view 3D reconstruction and camera pose estimation benchmarks demonstrate that FlashSSM-3D achieves competitive performance while requiring only ∼ 25% of the training data and less than 10% of the training compute used by VGGT. At inference time, FlashSSM-3D benefits from the practical efficiency of state space models, reconstructing scenes from 500+ views in under 10 seconds while maintaining competitive accuracy.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{shatwell2026georeferenced,

title = {Georeferenced Cross-View 3D Reconstruction from Ground and Satellite Images},

author = {David G Shatwell and Mubarak Shah},

url = {https://openreview.net/forum?id=h1zgNdJwcF

https://dshatwell23.github.io/documents/shatwell2026georeferenced.pdf},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Georeferenced 3D reconstruction from ground images requires recovering not only the local scene geometry, but also localizing every camera center and 3D point in a geographic coordinate frame. Existing cross-view geo-localization methods are able to estimate the GPS coordinate of a single ground camera, but do not reconstruct dense 3D geometry, while recent feed-forward multi-view 3D reconstruction models predict point maps and camera poses only in a local coordinate frame. We introduce X-VGGT, a feed-forward cross-view geometry transformer that unifies these tasks by jointly processing a set of ground-level images and a georeferenced satellite image to predict ground-view camera poses, dense 3D point maps, and a scene-level transform parameterized by gravity, scale, yaw and translation. This transform projects the local reconstruction into the satellite plane, resulting in the geo-registration of all ground images with the georeferenced satellite image. We further propose an evaluation protocol for this joint task and show that, across multiple cross-view datasets, X-VGGT produces accurate georeferenced reconstructions and outperforms prior cross-view geo-localization and multi-view 3D reconstruction models.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{bawatneh2026omnitom,

title = {OmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief Modeling},

author = {Adam Bawatneh and Sagar Sapkota and Amrit Singh Bedi and Santu Karmaker and Mubarak Shah},

url = {https://openreview.net/forum?id=wm4vgQFJmq

https://arxiv.org/abs/2605.26322},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Theory of Mind (ToM), the ability to infer others' knowledge, intentions, and emotions, is commonly evaluated in large language models (LLMs) using end-point question answering, where performance is judged solely by the final answer to a social reasoning query. This paradigm obscures whether the model actually constructs the underlying mental-state representations required for robust reasoning, particularly in scenarios involving divergent, evolving, or mistaken beliefs. In order to address this research gap, we introduce OmniToM, a benchmark that directly evaluates these representations by requiring explicit modeling of belief structures for all relevant actors within a narrative. These structures are composed of belief propositions: minimal statements of what an actor takes to be true about the world or another actor's mental state, allowing knowledge, intentions, emotions, and false beliefs to be analyzed in a common format. Models are evaluated in two stages: Stage 1: Belief Extraction, which extracts from the story the beliefs relevant to its social dynamics, and Stage 2: Belief Labeling, which assigns each belief a seven-dimensional schema label covering recursive order, truth status, knowledge access, explicitness, content type, mental source, and context. Built from 895 stories from the existing ToMBench story corpus and augmented with 22,343 labeled belief propositions, OmniToM uses a human-calibrated LLM-assisted annotation pipeline. Across diverse models in zero-shot evaluation, OmniToM reveals an actor-specific belief-tracking bottleneck: current LLMs struggle with the knowledge-access and representational decisions required to transform narrative facts into actors' beliefs and shared mental states.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{yousaf2026safe,

title = {Safe in Its Own Words: Self-Guided Safety Alignment for Multimodal Reasoning Models},

author = {Adeel Yousaf and Souradip Chakraborty and Mubarak Shah and Amrit Singh Bedi},

url = {https://openreview.net/forum?id=PFyto9A2fG

https://neurips.cc/virtual/2026/loc/atlanta/poster/152939},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Multimodal large reasoning models can reason over images and text, but they remain vulnerable to multimodal jailbreaks. Recent safety-alignment methods reduce attack success rate (ASR) by fine-tuning on reasoning traces from stronger teacher models. We show that this has a hidden cost of over-refusal. The aligned model learns to reject benign inputs that contain sensitive words or visual cues. This is especially harmful for boundary-safe inputs, where the prompt looks risky on the surface but is safe in context. We identify external teacher supervision as a key factor in this behavior. It introduces a distributional mismatch and shifts supervision away from the base model's own reasoning distribution. To address this, we propose SAGE (Safety-Aware Guided Elicitation), a self-guided data generation framework for multimodal safety alignment. SAGE guides generation in two stages: a reasoning-level guide first elicits a safety-aware trace, and a decision-level guide then elicits the final response conditioned on that trace. This factorized guidance lets SAGE shape both the reasoning process and the final safety decision without relying on external teacher traces. On LLaVA-CoT, SAGE reduces FigStep ASR from 86.2% to 2.1% and achieves an average over-refusal of 43.1, compared with 66.9--71.1 for prior reasoning-based safety-alignment methods, while preserving general utility. We will release the SAGE training dataset.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{lin2026efficient,

title = {Efficient and Transferable Agentic Knowledge Graph RAG via Reinforcement Learning},

author = {Junhong Lin and Shicheng Liu and Jinyeop Song and Song Wang and Julian Shun and Yada Zhu},

url = {https://arxiv.org/abs/2509.26383

http://github.com/junhongmit/KG-R1/},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Knowledge-graph retrieval-augmented generation (KG-RAG) couples large language models (LLMs) with structured, verifiable knowledge graphs (KGs) to reduce hallucination and provide reasoning traces. However, current KG-RAG systems often rely on fixed pipelines of multiple LLM modules (e.g., planning, reasoning, and responding), which inflate inference costs and tie performance to specific graph schemas. To address this, we introduce KG-R1, an agentic framework that optimizes KG-RAG through reinforcement learning (RL). Unlike modular workflows, KG-R1 uses a single agent that interacts with KGs as its environment, learning to retrieve information at each step and incorporating it into its reasoning and generation in a unified process. Across Knowledge-Graph Question Answering (KGQA) benchmarks, KG-R1 demonstrates both efficiency and transferability-using Qwen 2.5-3B, KG-R1 improves answer accuracy with fewer generation tokens than prior multi-module workflow methods that use much larger foundation or fine-tuned models. Furthermore, KG-R1 exhibits strong plug-and-play capability: after training, maintaining accuracy on unseen KGs without retraining. These properties make KG-R1 a promising KG-RAG framework for real-world deployment. Our code is publicly available at this http URL.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{zhang2026jarvis,

title = {JARVIS-Bench: Benchmarking Personal Intelligence Agents on Long-Horizon Real-User Daily Traces},

author = {Weizhi Zhang and Wei-Chieh Huang and Yueqing Liang and Liwei Jiang and Zhengxiang Wang and Liangwei Yang and Zechen Li and Yuchen Wu and Haozhen Zhang and Yu Wang and Yuanchen Bei and Yue Zhou and Siqi Zhu and Henry P Zou and Jiahong Liu and Xinni Zhang and Paul Martin and Joseph Marvin Imperial and Yuyang Luo and Xiongxiao Xu and Baixiang Huang and Shanglin Wu and Lucas Resck and Yibo Wang and Yuqing Liu and Langzhou He and Chengze Li and Yuxin Tian and Kening Zheng and Enze Ma and Boi Huynh and Zheng Hui and Rui Yang and Tao Feng and Jie Yang and Shanghao Li and Haoran Wang and Wooseong Yang and Hanrong Zhang and Huanhuan Ma and Daye Yoon and Yaozu Wu and Shikan Lian and Xiaoqian Ruan and Fangxin Wang and Hyeonjeong Park and Hins Hu and Wenzhe Fan and Chen Wang and Yifan Gu and Dongyuan Li and Song Wang and Aylin Caliskan and Jindong Wang and Xiao Luo and Yankai Chen and Kai Shu and Steve Liu and Philip S Yu},

url = {https://openreview.net/forum?id=yGQt9wAGkS},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Personal intelligence agents aim to support users across daily-life tasks by learning their preferences, routines, needs, and emotional states from long-term interaction. They are expected to anticipate user needs, provide proactive assistance, and infer preferences that users have never explicitly articulated. However, existing benchmarks for LLM personalization lack real user data spanning extended periods and diverse daily-life domains, and typically reduce personalization to simplified settings such as recommendation or fact retrieval. We introduce JARVIS-Bench, a new benchmark for evaluating personal intelligence agents, curated from real users’ daily traces and augmented with synthetic trajectory expansions grounded in implicit-preference insights from surveys and structured personas. JARVIS-Bench is built from participants who contributed 28 to 44 days of multimodal self-tracking logs, structured personas, and a 100-question implicit-preference survey spanning eight domains. It supports four downstream task surfaces: implicit preference reasoning, proactive-help prediction, emotion-aware modeling, and cross-domain transfer, all evaluated through a unified event iterator and long-horizon user-memory pipeline. We benchmark 28 LLMs from 10 model families, including Gemini, GPT, Qwen, Gemma, and Claude, across a five-condition input ladder ranging from no memory to lifelong memory. Results show that memory and model scaling improve personalization, yet current frontier models remain far from achieving robust real-world personal intelligence.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{liu2026mitigating,

title = {Mitigating Object Hallucination in Large Vision-Language Models via False Discovery Controlled Visual Data Splitting},

author = {Chang Liu and Yu Tian and Rui Xie},

url = {https://openreview.net/forum?id=ek5KY0D3CL

https://changliu1993-cl.github.io/CORAL/

https://arxiv.org/abs/2609.38979},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Multiple object hallucination, where large vision-language models (LVLMs) generate objects not supported by the visual input, is a persistent challenge caused by visual uncertainty during decoding. Existing methods reduce hallucinations using contrastive signals, but they rely on heuristics and lack principled control of false positives at the image level. To address this, we propose False Discovery Rate-COntRol of HALlucination (CORAL), a training-free framework that models visual uncertainty using an uncertainty-aware visual data splitting strategy and leverages mirror statistics to quantify visual contrast during decoding. By computing mirror statistics from paired, symmetrically perturbed visual inputs, CORAL estimates spurious object predictions and sets a data-driven threshold to control the expected fraction of false discoveries per image, suppressing hallucinations while retaining high power for truly grounded objects. The framework is flexible, supports multiple LVLMs, and mitigates hallucinations without retraining or supervision. Extensive experiments on multiple benchmarks with several evaluation metrics demonstrate that CORAL consistently outperforms state-of-the-art methods, providing more reliable and robust hallucination control. Code is available at: this https URL},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{li2026dd,

title = {DD-Ranking: Rethinking the Evaluation of Dataset Distillation},

author = {Zekai Li and Xinhao Zhong and Samir Khaki and Zhiyuan Liang and Yuhao Zhou and Mingjia Shi and Ziqiao Wang and Xuanlei Zhao and Wangbo Zhao and Ziheng Qin and Mengxuan Wu and Pengfei Zhou and Haonan Wang and David Junhao Zhang and Jia-Wei Liu and Shaobo Wang and Dai Liu and Linfeng Zhang and Guang Li and Kun Wang and Zheng Zhu and Zhiheng Ma and Joey Tianyi Zhou and Jiancheng Lv and Yaochu Jin and Peihao Wang and Kaipeng Zhang and Yiran Huang and Zhiwei Deng and Xindi Wu and George Cazenavette and Yuzhang Shang and Justin CUI and Jindong Gu and Qian Zheng and Hao Ye and Shuo Wang and Xiaobo Wang and Yan Yan and Angela Yao and Mike Zheng Shou and Tianlong Chen and Hakan Bilen and Baharan Mirzasoleiman and Manolis Kellis and Konstantinos N Plataniotis and Zhangyang ""Atlas"" Wang and Bo Zhao and Yang You and Kai Wang},

url = {https://arxiv.org/abs/2505.13300},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {In recent years, dataset distillation has provided a reliable solution for data compression, where models trained on the resulting smaller synthetic datasets achieve performance comparable to those trained on the original datasets. To further improve the performance of synthetic datasets, various training pipelines and optimization objectives have been proposed, greatly advancing the field of dataset distillation. Recent decoupled dataset distillation methods introduce soft labels and stronger data augmentation during the post-evaluation phase and scale dataset distillation up to larger datasets (e.g., ImageNet-1K). However, this raises a question: Is accuracy still a reliable metric to fairly evaluate dataset distillation methods? Our empirical findings suggest that the performance improvements of these methods often stem from additional techniques rather than the inherent quality of the images themselves, with even randomly sampled images achieving superior results. Such misaligned evaluation settings severely hinder the development of DD. Therefore, we propose DD-Ranking, a unified evaluation framework, along with new general evaluation metrics to uncover the true performance improvements achieved by different methods. By refocusing on the actual information enhancement of distilled datasets, DD-Ranking provides a more comprehensive and fair evaluation standard for future research advancements.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{buck2026topological,

title = {TopoGED: A Topological Encoder Decoder Framework for Temporal Graph Learning},

author = {Ronan Buck and Kiarash Shamsi and Tran Gia Bao Ngo and Astrit Tola and Baris Coskunuzer and Cuneyt Akcora},

url = {https://openreview.net/forum?id=cZEj9pgOP7},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{uddin2026nodes,

title = {From Nodes to Pixels: Topological and Structural Two-View Graph Imaging},

author = {Md Joshem Uddin and Soham Changani and Sai K Navuluru and Cuneyt Akcora and Baris Coskunuzer},

url = {https://github.com/sohamchangani/G2Image},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {G2Image is a deep learning framework that transforms graph-structured data into grid-based image representations. It leverages a dual-view approach—combining structural information (via Adjacency Grid Images) and topological features (via Multipersistence Toposurfaces)—to classify graphs using a CNN-Transformer architecture with Supervised Contrastive Learning.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{shamsi2026hydra,

title = {Hydra: Towards Transferable Multi-Task Learning on Temporal Graphs},

author = {Kiarash Shamsi and Farimah Poursafaei and Tran Gia Bao Ngo and Reihaneh Rabbany and Baris Coskunuzer and Guillaume Rabusseau and Michael Bronstein and Shenyang Huang and Cuneyt Akcora},

url = {

https://openreview.net/forum?id=prXn189GaI},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Real-world evolving networks are naturally modeled as temporal graphs (TGs), where capturing temporal dynamics is essential for predicting future graph properties that support downstream decision-making. Existing temporal graph methods have been developed primarily for single-task prediction, and little is known about their generalization across tasks or transfer to unseen networks. This leaves the challenge of multi-task graph property prediction in TGs largely open. We address this challenge by introducing Hydra, a novel architecture that integrates local connectivity features from temporal GNNs with a spectral learning module that captures global connectivity patterns. This design enables joint learning of local and global information under a multi-task objective. In multi-task classification, Hydra achieves an 8.9% relative gain in AUC over the strongest competitor. In multi-task regression, Hydra achieves competitive results in all three tasks, while obtaining the best results in two tasks with a 8.2% relative gain in MAE compared to the strongest baseline. Moreover, Hydra delivers these gains with a 22× reduction in training time compared to temporal transfer models. These results provide the first systematic evidence that multi-task transferable learning on temporal graphs is effective. By delivering consistent top-ranked performance, Hydra highlights multi-task training on temporal graphs as a promising direction toward adaptable foundation models for temporal graphs.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{zou2026conditional,

title = {Conditional Multi-Event Temporal Grounding in Long-Form Video},

author = {Yuanhao Zou and Arthad Kulkarni and Lucas Toñanez and Lincoln Spencer and Guangyu Sun and Tianxingjian Ding and Andong Deng and Yi Li and Shuangjun Liu and Yuan Li and Dashan Gao and Ning Bi and TAOTAO JING and Shuai Zhang and Chen Chen},

url = {https://openreview.net/forum?id=i6iM7B1Koa

https://arxiv.org/html/2606.15320v1},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Multimodal large language models have made rapid progress in video temporal grounding, yet real-world applications routinely require localizing every event that satisfies compositional temporal and spatial conditions. Existing benchmarks fall short: they localize only a single moment per query, count without temporal conditions, or treat grounding and counting as disjoint tasks. We introduce CoMET-Bench for Conditional Multi-Event Temporal Grounding in long-form video, comprising 2,789 queries over 600 videos averaging 33.8 minutes across five real-world domains, with each query composed from 4 temporal conditions, 3 spatial conditions, and a dedicated negative-query subset. We further propose a unified evaluation protocol jointly measuring counting, grounding, and negative-query recognition, including a new Rejection-F1 metric that prevents trivial gaming by lazy “always-empty” models. Benchmarking a broad suite of MLLMs, agent-based, and grounding-specialized methods reveals that existing approaches remain far from solving this task. Building on these findings, we propose CoMET-Agent, a training-free agentic framework that reformulates the task as structured search-and-aggregate, improving F1@0.5 by 6.1% over GPT-5 purely through structural reasoning. Failure analysis further surfaces three open directions: fine-grained entity tracking, position-uniform retrieval, and causal event pairing.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{kwon2026factorizedhmr,

title = {FactorizedHMR: A Hybrid Framework for Video Human Mesh Recovery},

author = {Patrick Kwon and Chen Chen},

url = {https://openreview.net/forum?id=7ERjBgSwbw

https://arxiv.org/abs/2605.14854},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Human Mesh Recovery (HMR) is fundamentally ambiguous: under occlusion or weak depth cues, multiple 3D bodies can explain the same image evidence. This ambiguity is not uniform across the body, as torso pose and root structure are often relatively well constrained, whereas distal articulations such as the arms and legs are more uncertain. Building on this observation, we propose FactorizedHMR, a two-stage framework that treats these two regimes differently. A deterministic regression module first recovers a stable torso-root anchor, and a probabilistic flow-matching module then completes the remaining non-torso articulation. To make this completion reliable, we combine a composite target representation with geometry-aware supervision and feature-aware classifier-free guidance, preserving the torso-root anchor while improving single-reference recovery of ambiguity-prone articulation. We also introduce a synthetic data pipeline that provides the paired image-camera-motion supervision under diverse viewpoints. Across camera-space and world-space benchmarks, FactorizedHMR remains competitive with strong baselines, with the clearest gains in occlusion-heavy recovery and drift-sensitive world-space metrics.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{dong2026rethinking,

title = {Rethinking Long-Video Efficiency: A Joint Allocation Perspective on Frames, Pixels, and Front-End Latency},

author = {Sixun Dong and Wei Li and Andong Deng and Qi Qian and Victor Zhu and Zhengping Ji and Chen Chen},

url = {https://sixundong.com/projects/lohi},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Efficient long-video understanding with vision-language models (VLMs) is often framed as selecting informative frames or visual tokens at a fixed native resolution. We show that per-frame resolution can instead be traded for denser temporal coverage, while front-end decoding latency depends on the size of the candidate pool rather than the final token budget. An empirical study across multiple VLMs and long-video benchmarks yields three findings: dense low-resolution sampling outperforms sparse native-resolution sampling at matched token budgets; resolution-sensitive tasks benefit from selected high-resolution frames; and front-end decoding dominates wall time for hour-long videos. Motivated by these findings, we introduce LoHi, a training-free, single-pass framework that combines a dense low-resolution video stream with sparse high-resolution image frames through the VLM's native video and image pathways. LoHi-Anchor selects high-resolution frames using codec I-frame metadata, while LoHi-SemDiv uses query relevance and visual diversity over CLIP features. Across three long-video benchmarks, LoHi improves average accuracy by 10.6 percentage points over the native-resolution baseline at a matched token budget and by 5.2 percentage points over the strongest prior efficiency method. It also reduces front-end decoding latency by up to 7x on hour-long videos. Project page: this https URL},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{kowsher2026where,

title = {Where You Backpropagate Matters: Token Hypothesis for Memory-Efficient Fine-Tuning},

author = {Md Kowsher and Chen Chen},

url = {https://openreview.net/forum?id=8D8sa1Nuh0},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Fine-tuning large transformers routes gradients through every token position, paying the cost in activation memory that scales linearly with sequence length and dominates the GPU budget. Existing efficient methods reduce the parameter side—low-rank adapters, prefix tuning, weight-decomposed updates—but leave the activation footprint untouched. We present the Token Hypothesis, a memory-efficient fine-tuning procedure motivated by a Neural Tangent Kernel analysis of how fine-tuning signal is distributed across token positions. We prove that every position is a low-dimensional capacity bottleneck whose width does not grow with the model, that positions with aligned hidden-state geometry are interchangeable in bidirectional models, and that causal positions form a monotone capacity chain in which later positions strictly dominate earlier ones. The analysis provides an assumption-qualified criterion for estimating how many positions are needed and which ones, using quantities computable from forward passes alone: in bidirectional models, small arbitrary subsets are predicted to suffice when position-wise kernels are aligned, while in causal models, later positions are predicted to carry the richest fine-tuning signal. We translate this into a procedure that backpropagates through only a small subset of positions while running the full forward pass over the rest. Empirically, the procedure matches or exceeds full-sequence fine-tuning across six task families—commonsense reasoning, mathematical reasoning, visual question answering, video question answering, GLUE, and image classification—and five backbones spanning bidirectional and causal architectures. It reduces activation memory substantially, translating to up to ~20% lower total GPU memory (~33 GB) on long-context tasks, and composes cleanly with parameter-efficient methods such as LoRA and DoRA.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{lyu2026mls,

title = {MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI},

author = {Bohan Lyu and Yucheng Yang and Siqiao Huang and Jiaru Zhang and Qixin Xu and Xinghan Li and Xinyang Han and Huaqing Zhang and Yicheng Zhang and Runhan Huang and Kaicheng Yang and Zitao Chen and Wentao Guo and Junlin Yang and Xinyue Ai and Wenhao Chai and Yadi Cao and Ziran Yang and Kun Wang and Dapeng Jiang and Huan-ang Gao and Shange Tang and Chengshuai Shi and Simon Du and Max Simchowitz and Jiantao Jiao and Dawn Song and Chi Jin},

url = {https://openreview.net/forum?id=EaF7FneBfZ

https://mls-bench.com/},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Modern AI progress has been driven by ML methods that are generalizable across settings and scalable to larger regimes. As large language models demonstrate advanced capabilities in reasoning, coding, and engineering tasks, it is increasingly important to understand whether they can discover such methods rather than only apply existing ones. We introduce MLS-Bench, a benchmark for evaluating whether AI systems can invent generalizable and scalable ML methods. MLS-Bench contains 140 tasks across 12 domains, each requiring an agent to improve one targeted component of an ML system or algorithm and demonstrate that the improvement generalizes across controlled settings and scales. We find that current agents remain far from reliably surpassing human-designed methods, and that engineering-style tuning is easier for them than genuine method invention. We further study the effects of test-time scaling, adaptive compute allocation, and context provision on agents' discovery performance, together with case studies of their behavior. Our analyses suggest that the bottleneck is not only in proposing new methods, but also in the scientific insight needed to plan, validate, and scale claims about them. More search, compute, or context alone does not remove this bottleneck. We build and maintain a community platform for cumulative and comparable iteration, and release the data and code at this https URL.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{jore2026bifurcation,

title = {Bifurcation Models: Learning Set-Valued Solution Maps with Weight-Tied Dynamics},

author = {Caleb Jore and Jialin Liu},

url = {https://openreview.net/forum?id=tVZohLsVHC},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Many scientific and combinatorial problems admit multiple correct solutions, not a single label. Standard supervised learning resolves this ambiguity by choosing one solution as the target, but this hidden selector can be arbitrary, discontinuous, and harder to learn than the underlying solution set. We study bifurcation models, a weight-tied dynamical view in which different initializations can converge to different stable equilibria, so the model represents an attractor landscape rather than one chosen branch. We prove that broad set-valued maps with locally Lipschitz branches can be represented by regular equilibrium dynamics and that the induced selectors are almost everywhere regular, while manual selectors can be arbitrarily irregular. Experiments on frustrated Ising models show that such dynamics can discover multiple valid equilibria without branch labels and outperform single-branch supervision. Allen--Cahn experiments further show that diversity is not automatic: it can be encouraged explicitly, but with an accuracy--diversity tradeoff.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{do2026simulating,

title = {Simulating Students or Sycophantic Problem Solving? On Misconception Faithfulness of LLM Simulators},

author = {Heejin Do and Shashank Sonkar and Mrinmaya Sachan},

url = {https://openreview.net/forum?id=t2Q9ciLFrV

https://arxiv.org/abs/2605.12748},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Large language models (LLMs) can fluently generate student-like responses, making them attractive as simulated students for training and evaluating AI tutors and human educators. Yet such simulators are typically evaluated by output similarity to real students, not by whether they behave like students with coherent misconceptions during interaction. We introduce a controlled framework for evaluating misconception faithfulness, whether a simulator maintains a misconception-driven belief state and updates selectively when feedback addresses the underlying misconception. Central to our framework is a misconception-contrastive feedback protocol that compares targeted feedback against two controls: misaligned feedback (targeting a different but plausible misconception) and generic feedback (only identifying answer is wrong). We propose Selective Flip Score (SFS), which quantifies how much more often a simulator flips its answer under targeted feedback than under contrastive controls. Across seven LLMs (4B-120B), multiple datasets, and prompting strategies, simulators exhibit near-zero SFS, correcting their answers at similarly high rates regardless of feedback relevance. Further analyses reveal a sycophantic failure mode: models behave less like students with misconceptions but more like problem-solvers who treat any corrective signal as a cue to abandon the simulated belief and re-solve from internal knowledge. To address this, we develop a post-training pipeline spanning supervised fine-tuning (SFT), preference optimization, and reinforcement learning (RL) with an SFS-aligned reward; SFT yields notable gains up to +0.56, and SFS-aligned RL provides more consistent improvements than preference optimization. Our results establish misconception faithfulness as a challenging yet trainable property, motivating a shift from static output matching toward interactive, belief-aware student modeling.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{ahmed2026robust,

title = {Robust Nash Alignment under Preference Uncertainty},

author = {Shihab Ahmed and Debamita Ghosh and David Tang and Yudan Wang and Alvaro Velasquez and Yue Wang},

url = {https://openreview.net/forum?id=CiLFISXIZl},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Preference-based alignment methods typically optimize against a single preference model, and can therefore be brittle when pairwise preferences are uncertain: noisy, heterogeneous, or shift after deployment. To address these issues, we propose Robust Nash Alignment, a game-theoretic framework for alignment to uncertain pairwise preferences. Our formulation has a major learner seeking a policy with a large worst-case win rate against both an adversarial competitor and any preference kernel lying in an ambiguity set around a nominal preference. When the ambiguity set captures the uncertainty in preferences, the resulting robust objective of the game directly yields a certified lower bound on worst-case performance. However, we note this problem is computationally challenging to optimize, and to address this, we introduce a four-player primal-dual proxy game involving the leader policy, follower policy, adversarial kernel, and dual variable, and develop a single-loop optimistic mirror descent-ascent algorithm for it. We show that the proxy always lower-bounds the truncated hard-constrained objective, quantify the proxy-to-hard gap, and characterize an exactness condition under which the proxy recovers the robust objective. We then prove an (mathcal{O}(1/sqrt{T})) average-iteration convergence for the proxy-game duality gap, which implies a near-optimal robust policy for the original robust objective. Experiments on controlled tabular games and LLM alignment with uncertain preference further validate the convergence theory and show improved performance over nominal baselines.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{sklaviadis2026bayesian,

title = {The Bayesian Learning Rule Beyond KL Geometry},

author = {Sophia Sklaviadis and Wu Lin and Mario A T Figueiredo and Andre Martins and Thomas Möllenhoff and Mohammad Emtiyaz Khan},

url = {https://openreview.net/forum?id=iMEeZOdaUi

https://arxiv.org/abs/2610.31308},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {The Bayesian learning rule (BLR) exploits the relation between Bayesian updating and the Kullback-Leibler divergence (KLD) to derive efficient additive updates for exponential family variational posteriors. The KL geometry is powerful but restrictive: combined with posterior families with parameter-dependent support it can lead to undefined objectives. We ask which non-KL geometries preserve the additive BLR computation. Previous works that use f-divergences suggest that this is only possible with the KLD. Here, we show that this is not true. We show that additive information-geometric optimization algorithms can be derived for the class of mixed-space Bregman or Fenchel-Young discrepancies. These discrepancies induce generalized maximum-entropy families, dual coordinates, and generalized-BLR canonical gradient updates, recovering the classical BLR. The generalized BLR leads to new algorithms that textit{cannot} be derived from the KL-based BLR. For the Tsallis negentropy with quadratic statistics, the generalized BLR yields bounded-support and heavy-tailed q-Gaussian posteriors and an easy Newton-like implementation. Experiments up to CIFAR-100 illustrate controllable support and tail behavior and competitive performance.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{minut2026soap,

title = {SOAP-Bubbles: Effective and Scalable Variational Learning with Structured Covariances},

author = {Adrian Robert Minut and Nico Daheim and Marco Miani and Mohammad Emtiyaz Khan and Wu Lin and Thomas Möllenhoff},

url = {https://openreview.net/forum?id=YBJkwSJVbo

https://arxiv.org/abs/2606.23357},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Structured weight-uncertainty can improve many aspects of deep learning, but it remains costly to estimate and difficult to implement. Here, we show that these issues can be addressed by adapting the SOAP optimizer. Our key idea is to run IVON, an existing diagonal-covariance variational method, in the eigenspace of SOAP's preconditioner and then use the preconditioner to transform the diagonal estimate into a non-diagonal covariance. The resulting method has costs similar to those of SOAP and requires no drastic changes to training pipelines. We call the posteriors obtained in this way SOAP-Bubbles and our new optimizer Eigenspace-VON (EVON). We show that, for logistic regression, EVON recovers the exact Gaussian covariance and that, for language model pretraining, it yields significantly better results than existing diagonal-covariance methods. Our work makes it easier to estimate more expressive posterior distributions for deep learning at scale.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{persaud2026brute,

title = {Brute-Force Jailbreaks and Codon-Aware Watermarking for DNA Foundation Models},

author = {Munish B Persaud and Amrit Singh Bedi and Souradip Chakraborty},

url = {https://openreview.net/forum?id=6JUgH59aJ1

https://neurips.cc/virtual/2026/loc/atlanta/poster/155344},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{chehade2026test,

title = {Test-time Risk Adaptation with Mixture of Agents},

author = {Mohamad Chehade and Amrit Singh Bedi and Souradip Chakraborty and Amy Zhang and Hao Zhu},

url = {https://openreview.net/forum?id=xjpWCBDACk},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Deployed reinforcement learning agents must satisfy safety requirements that emerge only at test time—evolving regulations, unexpected hazards, or shifted operational priorities. Current risk-aware methods embed fixed risk models (typically return variance) during training, but this approach suffers from two fundamental limitations: it restricts risk expressiveness to trajectory-level statistics, and it induces uniform conservatism that reduces behavioral coverage needed for effective deployment adaptation. We propose TRAM (Test-time Risk Adaptation via Mixture of Agents), a deployment-time framework that composes risk-neutral source policies to satisfy arbitrary risk specifications without retraining. TRAM represents risk through occupancy-based functionals that capture spatial constraints, behavioral drift, and local volatility—risk types that trajectory variance cannot encode. Our theoretical analysis provides localized performance bounds that cleanly separate reward transfer quality from risk alignment costs, and proves that risk-neutral source training is minimax optimal for deployment risk adaptation. Empirically, TRAM delivers superior safety-performance trade-offs across gridworld, continuous control, and large language model domains while maintaining computational efficiency through successor feature implementation.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{barakat2026why,

title = {Why Pass@k Optimization Can Degrade Pass@1: Prompt Interference in LLM Post-Training},

author = {Anas Barakat and Souradip Chakraborty and Khushbu Pahwa and Amrit Singh Bedi},

url = {https://openreview.net/forum?id=TCynOWswES

https://arxiv.org/abs/2602.21189},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Pass@k is a widely used performance metric for verifiable large language model tasks, including mathematical reasoning, code generation, and short-answer reasoning. It defines success if any of k independently sampled solutions passes a verifier. This multi-sample inference metric has motivated inference-aware fine-tuning methods that directly optimize pass@k. However, prior work reports a trade-off: pass@k improves while pass@1 degrades under such methods. This trade-off is practically important because pass@1 often remains a hard operational constraint due to latency and cost budgets, imperfect verifier coverage, and the need for a reliable single-shot fallback. We study the origin of this trade-off and provide a theoretical characterization of when pass@k policy optimization can reduce pass@1 through gradient conflict induced by prompt interference. This interference arises because shared model parameters couple prompts, so improving performance on some prompts can adversely affect others. We show that pass@k policy gradients can conflict with pass@1 gradients because pass@k optimization implicitly reweights prompts toward low-success prompts; when these prompts are negatively interfering, their upweighting can rotate the pass@k update direction away from the pass@1 direction. We illustrate our findings with large language model experiments on verifiable mathematical reasoning and coding tasks.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{cui2026closer,

title = {A Closer Look at Dynamic Scene Graph Generation in the Era of Multimodal Large Language Models},

author = {Xuanming Cui and Jaiminkumar A Bhoi and Chionh W Peng and Adriel Kuek and Ser Nam Lim},

url = {https://openreview.net/forum?id=Y1D9E4ZhwT

https://arxiv.org/abs/2503.15846},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

abstract = {Dynamic Scene Graph Generation (DSGG) aims to capture objects and their evolving relations in videos. Despite recent progress, the practicality and quality of generated scene graphs remain limited compared to the rapid advances in Multimodal Large Language Models (MLLMs). In this work, we revisit DSGG from two fundamental perspectives: task setup and model design. From the task setup perspective, we identify two key limitations of the current recall-oriented evaluation protocol: (i) a severe precision-recall trade-off, and (ii) uninformative and redundant relation generation. To better assess practical usefulness, we introduce five additional metrics that provide a more comprehensive evaluation of the quality of generated dynamic scene graphs. From the model design perspective, we explore directly using MLLMs for scene graph generation and establish a strong MLLM-based DSGG baseline through three design changes. First, we replace the conventional bottom-up pipeline with a top-down reason-then-locate strategy. Second, we reformulate frame-wise dynamic graphs as Temporal Relation Set (TRS) prediction, improving both efficiency and performance. Third, we introduce Importance-Aware Finetuning (IAF) to encourage more relevant and diverse relation generation. Extensive experiments on Action Genome, VidVRD, and PVSG show that our approach consistently achieves state-of-the-art performance.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{hossain2026evaluating,

title = {Evaluating Whether LLMs Can Reliably Connect the DOTs?},

author = {Eftekhar Hossain and John Salvador and Santu Karmaker},

url = {https://openreview.net/forum?id=St4No81OAc},

year  = {2026},

date = {2026-01-01},

urldate = {2026-01-01},

booktitle = {Advances in Neural Information Processing Systems},

publisher = {NeurIPS 2026},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}