Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents

RL post-training yields progress advantage, a log-ratio that recovers optimal step-level advantage without dedicated reward models, outperforming trained alternatives across agent benchmarks.

Changdae Oh, Wendi Li, Seongheon Park, Samuel (Min-Hsuan) Yeh and 2 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 13 on Hugging Face · Code ★ 12

Distributionally Robust Multi-Task Reinforcement Learning via Adaptive Task Sampling

Nicholas Corrado, Wenyuan Huang, Josiah Hanna

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

When Does Knowing the State Help? Diagnosing Process vs. Outcome Reward Design

Wenpei Shao, Ross Jacobucci

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

Spectral-Spatial Interpretation

Haotian Ma, Ruqi Yang, Philip Townsend

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

Complementary Cache Guidance with Gradient Disentanglement for Continuous Test-Time Adaptation

Fanchun Meng, Yuhang Pei, Jiazhen Huang, Tao Ren and 3 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

SKIM: Pruning Large Language Model Agents via Selective Knowledge Informed Masking

Moonseok Choi, Giung Nam, Jongwon Jeong, Minki Kang and 1 more

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

Beyond Global Alignment: Structured Compositional Reasoning for Vision-Language Models

Zhoujun Ye, Yiwei Fu, Qiyun Huang, Jie Yang and 2 more

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

Fair Division of Work in Collaborative Mean Estimation via Bargaining

Michael O. Harding, Alex Clinton, Kirthevasan Kandasamy

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

Approximation Algorithms for GPU Pricing under Finite Capacity

Yaolong Yu, Hanrui Zhang, Zeyu Zheng, Kirthevasan Kandasamy

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

Uncovering Semantic Hierarchies in Text-Attributed Graphs via Variational EM-based LLM–GNN Synergy

Yunhui Liu, Xudong Jin, Qizhuo Xie, Chunhui Zhao and 4 more

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

PATH: A Dual Perspective for High-quality Text-attributed Graph Learning

Yuhang Pei, Fanchun Meng, Changhu Wang, Tao Ren and 5 more

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

CLEAR: Complementary Tripartite Play with Bayesian Calibration for Semi-Supervised Edge Classification

Zhipeng Sun, Fanchun Meng, Jiazhen Huang, Yongpeng Zhang and 4 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

AnaDiffusion: Anatomically Compositional Latent Diffusion for Controllable 3D Brain MRI Generation

Tracy Han, Lulin Liu, Bangya Liu, Yuanhao Cai and 7 more

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

DEBATE: A Large-Scale Benchmark for Evaluating Opinion Dynamics in Role-Playing LLM Agents

Yun-Shiuan Chuang, Ruixuan Tu, Chengtao Dai, You Li and 7 more

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

SMI: Semantic Medical ID for Hierarchy-Aware Concept Representation

Ziyang Song, Lia Shen, Yixuan Li, Qincheng Lu and 5 more

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

LAPrune: Logits-Aligned Scoring Proxy for KV Pruning via Vector Quantization

Mingyang Yu, Rong-Cheng Tu, Yifu Ding, Hanqing Zhao and 3 more

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

Anatomy of Off-Policy Policy Gradient: Importance Sampling, KL Regularization, and Baselines

Haoqun Cao, Yurun Yuan, Tengyang Xie

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Entropy Dynamics of Agent Reinforcement Learning

Wendi Li, Shawn Im, Sharon Li

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

Provably Efficient Representation Learning for Low-Rank CMDPs

Kaixuan Liu, GUOJUN XIONG, Shengpu Tang, Wanyun Si and 1 more

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

Fast Sandwich Products in Clifford Algebra

Travis Pence, Daisuke Yamada, Jiaqi Mo, Chanyoung Moon and 2 more

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

Learning Optimal Transport Plans Via Autoregressive Token Regression

Ivan J Marquez, Takis Chytas, Vikas Singh

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

Noise-Level KL Rates for Multi-Marginal Schrödinger Bridge Surrogates

Hui Chen, Shen Xu, Vikas Singh

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Capturing LLM Capabilities via Evidence-Calibrated Query Clustering

ECC calibrates semantic embeddings with limited model comparisons to cluster queries by latent capability demands, improving LLM ranking by ~18 points over semantic baselines and aiding query routing.

Fangzhou Wu, Sandeep Silwal, Qiuyi (Richard) Zhang

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face

NPUsper: Eliminating Redundant Computation for Real-Time Whisper on Mobile NPUs

NPUsper eliminates redundant Whisper computation on mobile NPUs via online hallucination detection and chunked decoding to cut latency, TTFT, and power.

Hojeong Lee, Si H Lee, Sungwon Woo, Chengpo Yan and 2 more

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

PG-LRF: Physiology-Guided Latent Rectified Flow for Electro-Hemodynamic PPG-to-ECG Generation

PG-LRF uses a physiology-guided latent rectified flow with an electro-hemodynamic simulator to generate physiologically plausible ECGs from PPG, improving generation and cardiovascular disease classification.

Xiaoda Wang, Minxiao Wang, Kaiqiao Han, Defu Cao and 9 more

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

Fisher-R1: Training LLM Agents for Reliable Hypothesis Testing

P-Bench reveals LLM agents make subtle inferential errors in hypothesis testing, and Fisher-R1 improves reliability via reinforcement learning to outperform GPT-5.4 and DeepSeek-V4-Pro.

Jiacheng Miao, Jin Mu, Guanhua Chen, James Zou

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face

Online Localized Conformal Prediction

OLCP combines online adaptation with covariate localization for efficient online conformal prediction under heterogeneity, with OLCP-Hedge selecting bandwidths via online expert aggregation; both achieve valid long-run coverage with narrower prediction sets.

Yuheng Lai, Garvesh Raskutti

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

Shattered Compositionality: Counterintuitive Learning Dynamics of Transformers for Arithmetic

Transformers learn arithmetic skills non-sequentially via correlational matching, causing mixing errors and poor robustness to distribution shifts even when scaled.

Xingyu Zhao, Darsh Sharma, Rheeya Uppaal, Yiqiao Zhong

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

ROCKET: Residual-Oriented Multi-Layer Alignment for Spatially-Aware Vision-Language-Action Models

ROCKET aligns multiple VLA layers to a 3D vision model via residual streams and shared projectors, achieving near-state-of-the-art LIBERO success with about 4% compute.

Guoheng Sun, Tingting Du, Kaixi Feng, Chenxiang Luo and 5 more

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Generating the Unheard: Phylogeny-Guided Latent Generation for Ancestral Sound Reconstruction

This framework generates ancestral bird vocalizations by inferring decodable VAE latents guided by phylogenetic traits, achieving genuine generation and naturalistic audio quality.

Tianyi Xu, Shrinaath Narasimhan, Evan Gorstein, Santiago Perea and 2 more

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

CoMMa: Contribution-Aware Medical Multi-Agents for Decentralized Oncology Decision Support

CoMMa is a decentralized multi-agent oncology framework using partitioned data, specialized fine-tuning, and deterministic contribution-aware aggregation to enable interpretable, privacy-preserving decision support across diverse clinical settings.

Yichen WU, Yujin Oh, Sangjoon Park, Kailong Fan and 9 more

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

MAST: Label-Efficient, Robust, and Generalizable Sound Detection for Biodiversity Monitoring via Masked Audio Pretraining and Self-Training

MAST combines masked audio pretraining and self-training to improve sound detection across ecological domains, achieving substantial cross-site gains with minimal labeled data.

Tianyi Xu, Daniel Pimentel-Alarcón, Zuzana Buřivalová, Claudia Solis-Lemus

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

Prudent-Banker: No Extra Fees for Baseline Safety in Adversarial Bandits With and Without Delays

Prudent-Banker achieves minimax adversarial bandit regret with near-constant safe baseline regret despite arbitrary delayed feedback, matching new lower bounds.

Ting Hu, Luanda Cai, Emmanouil-Vasileios Vlatakis-Gkaragkounis

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

Learning Visual Feature-Based World Models via Residual Latent Action

Residual Latent Action predicts visual feature dynamics via flow matching, outperforming diffusion world models with orders-of-magnitude faster inference and enabling offline robot learning from videos.

Xinyu Zhang, Zhengtong Xu, Yutian Tao, Yeping Wang and 2 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 3 on Hugging Face · Code ★ 47

Task Vector Geometry Underlies Dual Modes of Task Inference in Transformers

Task-vector geometry governs dual inference: in-distribution tasks use convex combinations of learned vectors, while out-of-distribution tasks use nearly orthogonal extrapolative subspaces.

Hao Yan, Haolin Yang, Yiqiao Zhong

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

Testable Learning of General Halfspaces under Massart Noise

A testable learning algorithm learns general Massart halfspaces under Gaussian marginals with quasi-polynomial complexity matching SQ lower bounds.

Ilias Diakonikolas, Giannis Iakovidis, Daniel Kane, Sihan Liu

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Polynomial-Time Robust Multiclass Linear Classification under Gaussian Marginals

Multiclass linear classification under Gaussian marginals achieves polynomial-time robust learning via pairwise and localization frameworks, yielding near-optimal error bounds and exposing perceptron limitations.

Ilias Diakonikolas, Giannis Iakovidis, Mingchen Ma

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

Hide-and-Seek in Trajectories: Discovering Failure Signals for VLA Runtime Monitoring

Hide-and-Seek formulates VLA failure detection as coarsely supervised learning to localize failure-indicative actions from trajectory-level labels alone via contrastive objectives, achieving state-of-the-art multi-task detection with practical accuracy-timeliness trade-offs.

Seongheon Park, Wendi Li, Changdae Oh, Samuel (Min-Hsuan) Yeh and 3 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 8 on Hugging Face

Multi-Head Recurrent Memory Agents

Multi-Head Recurrent Memory partitions recurrent agent memory into independent heads to prevent overwriting, boosting long-context retention from under 30% to 74% at 896K tokens.

Jiatong Li, Samuel (Min-Hsuan) Yeh, Sharon Li

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

Tracing Agentic Failure from the Flow of Success

OAT trains neural controlled differential equations on successful agent trajectories to detect failure steps without failure annotations, outperforming prompting baselines by up to 20% F1 with 200-5000x speedup.

Samuel (Min-Hsuan) Yeh, Yiwen Zhu, Shaleen Deep, Sharon Li

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 16 on Hugging Face

Corrective Diffusion Language Models

Standard diffusion language models lack reliable token correction, so a correction-oriented post-training principle improves iterative refinement and outperforms masked diffusion baselines.

Shuibai Zhang, Fred Peng, Yiheng Zhang, Jin Pan and 1 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face · Code ★ 17

Finding Koopman Invariant Subspaces via Personalized PageRank

Personalized PageRank detects Koopman-invariant dictionary subspaces via EDMD zero blocks with finite-sample guarantees and controls multi-step leakage without assuming invariance.

Hyukpyo Hong, Qin Li, Matthew J Colbrook, Hanbaek Lyu

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Optimizing Computational-Statistical Runtime for Wasserstein Distance Estimation

<|message_model|><|content_text|>A Sample-Sketch-Solve paradigm estimates squared Wasserstein distance between smooth distributions within ε error in near-optimal time. For α-Hölder smooth distributions on (0,1)^d it achieves ε^{-max(2,(d+1+o(1))/(1+α))} runtime, with optimal Θ(ε^{-2}) in 2D for α>1

Peter M Jacobs, Jeff M Phillips

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

KV-PRM: Efficient Process Reward Modeling via KV-Cache Transfer for Multi-Agent Test-Time Scaling

KV-PRM eliminates text re-encoding by scoring via pre-existing KV caches, reducing process reward modeling cost from quadratic to linear and cutting latency and FLOPs by orders of magnitude.

Peng Kuang, Haibo Jin, Xiaoyu Han, Yanli Wang and 4 more

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

No Coin Left Behind: Maximizing Strategic Surplus Against No-Regret Dynamics

Against no-regret FTRL learners, strategic surplus scales with suboptimal actions and game randomness, revealing a geometric dichotomy between steep and non-steep regularizers.

Yiheng Su, Emmanouil-Vasileios Vlatakis-Gkaragkounis

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