False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents

Self-evolving search agents suffer co-cheating where proposers and solvers mutually reinforce errors; CrossFit partitions sources to cross-fit agreement and cuts false agreement by over half, boosting downstream search by 8+ points.

Meijia Chen, Hao Li, Zheng Lu, Hongshan Lin and 11 more

Published Sep 30, 2026 · 0 citations · ▲ 670 on Hugging Face

Video2Skill: From Streaming Experience to Reusable Embodied Skills

Video2Skill benchmarks streaming embodied skill discovery, showing VLMs group manipulation events poorly and rarely expand skill libraries despite supervised fine-tuning.

Jianshu Zhang, Ce Zhang, Xiyuan Yang, Chenwei Xu and 5 more

Published Sep 29, 2026 · 0 citations · ▲ 11 on Hugging Face

WM-VLM: Probing Internal World Models for Interleaved Visual-Textual Reasoning

WM-VLM adds a world-model branch to vision-language models to generate intermediate visual states for spatial reasoning, outperforming baselines by up to 39.25 points on mental rotation tasks.

Yuheng Zha, Yilei Wang, Qiyue Gao, Junrong Chen and 4 more

Published Sep 28, 2026 · 0 citations · ▲ 11 on Hugging Face · Code ★ 1

A Survey of Agentic Reasoning for Large Language Models: Towards Recursively Self-Improving and Collective Agents

This survey organizes LLM agentic reasoning into foundational, self-evolving, and collective layers, distinguishing in-context and post-training methods across applications while outlining open challenges.

Tianxin Wei, Ting-Wei Li, Zhining Liu, Xuying Ning and 25 more

Published Jan 18, 2026 · 1 citation · ▲ 208 on Hugging Face · Code ★ 1,398

Efficient Memory Management for Large Language Model Serving with PagedAttention

PagedAttention applies OS virtual memory and paging to LLM key-value caches, reducing waste and duplication to boost vLLM throughput 2-4x over existing systems.

Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng and 5 more

Published Sep 12, 2023 · 50 citations · ▲ 76 on Hugging Face · Code ★ 86,094

PAR: Political Actor Representation Learning with Social Context and Expert Knowledge

PAR learns political actor representations by integrating social context and expert knowledge into a unified framework.

Shangbin Feng, Zhaoxuan Tan, Zilong Chen, Ningnan Wang and 4 more

Published 2022 · 8 citations

Compressible Representations: Functional Spines in Deep Neural Networks

Niranjan Rajesh, Meenakshi Khosla

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Reasoning Pathologies in Large Language Models: A Diagnostic Perspective

Rohan Surana, Junda Wu, Sheldon Yu, Gagan Mundada and 5 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

(Strongly) Replicable Distribution Testers imply High Probability Distribution Testers

Ilias Diakonikolas, Jingyi Gao, Daniel Kane, Sihan Liu and 1 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

CyCLeGen: Cycle-Consistent Layout Prediction and Image Generation

Shan Xiaojun, Haoyu Shen, Yucheng Mao, Haiyang Xu and 4 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Evaluating Spatiotemporal Reasoning of Vision-Language Models in Atari Gameplay

Mingjia Huo, Yao Fu, Bo Chang, Yaqing Wang and 7 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Gumbo: Gumbel Optimized High-Temperature Speculative Sampling

Jonah Yi, Dan Fu, Yu-Xiang Wang

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Distance Marching for Generative Modeling

Distance Marching improves time-unconditional generative models via distance-focused losses and inference, surpassing flow matching FID with fewer steps and aiding OOD detection.

Zimo Wang, Ishit Mehta, Haolin Lu, Xunpeng Huang and 4 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

TAC: Timestamped Audio Captioning

TAC generates temporally grounded audio captions via synthetic training, reducing hallucinations and outperforming competitors in detection and dense captioning; cascading it with LLMs achieves state-of-the-art audio and audio-visual reasoning.

Sonal Kumar, Prem Seetharaman, Ke Chen, Oriol Nieto and 7 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

3DSPA: A 3D Semantic Point Autoencoder for Evaluating Video Realism

3DSPA is a reference-free 3D semantic point autoencoder that evaluates video realism by integrating point trajectories, depth, and semantic features to detect physical violations and motion artifacts with high human alignment.

Bhavik Chandna, Kelsey Allen

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Wavelet Flow Matching for Multi-Scale Physics Emulation

Wavelet Flow Matching performs optimal-transport in multi-scale wavelet space via U-Net velocity prediction for stable, accurate generative PDE emulation without autoencoders.

Gabriele Accarino, Juan Nathaniel, Carla Roesch, Pierre Gentine and 3 more

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

AgentKVShift: Efficient KV Cache Reuse for Agentic Memory Systems

AgentKVShift uses probe-guided KV residual correction to reuse agentic memory caches with near-full accuracy at 10-30% recompute, yielding 2-3.5x prefill speedups.

Nilesh Pandey, Jason Kong, Lanxiang Hu, Quanling Zhao and 4 more

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Calibrating LLMs with Semantic-level Reward

Calibration with Semantic Reward improves LLM calibration by replacing token-level confidence with direct semantic-space rewards, reducing ECE by up to 40% and raising AUROC by up to 31%.

Fengfei Yu, Ruijia Niu, Dongxia Wu, Yian Ma and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

TIER: Trajectory-Invariant Execution Rewards for Multi-Step Tool Composition

TIER derives dense tool-use rewards from execution and schemas rather than reference paths, enabling over 90% accuracy on multi-step composition where trajectory supervision fails.

Anay Kulkarni, Chia En Lu, Dheeraj Mekala, Jayanth Srinivasa and 2 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Language-Based Agent Control

LBAC is a programming model that enforces user policies on agentic applications by requiring agents to generate well-typed programs rejected by a type-checker before execution.

Timothy Zhou, Loris D'Antoni, Nadia Polikarpova

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026