Cross-Lingual Alignment for Decoder-Only Models using MoE Routers

Cross-lingual MoE router alignment improves multilingual LLM performance by aligning router outputs across languages instead of hidden states.

Lucas Bandarkar, Clark Peng, Ahmed Haj Ahmed, Aditi Khandelwal and 1 more

Published Oct 1, 2026 · 0 citations · ▲ 1 on Hugging Face · Code

Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems

SciUtopia is a closed-loop LLM simulation framework modeling entire academic ecosystems; it finds resubmission amplifies reviewer burden, cautious exploration balances impact and diversity, and inequality can emerge without cumulative funding advantage.

Yiqiao Jin, Yiyang Wang, Lucheng Fu, Bing He and 7 more

Published Oct 1, 2026 · 0 citations · ▲ 40 on Hugging Face · Code ★ 29

Multilinguality in Hybrid Attention LLMs

Hybrid attention LLMs develop cross-lingual alignment tied to recurrent and full-attention layer ordering, with a spike at the first full-attention layer; distillation shows starting with full attention learns up to 2.5× faster.

Lucas Bandarkar, Junlin Hu, Chenyuan Yang, Mohsen Fayyaz and 1 more

Published Sep 28, 2026 · 0 citations · ▲ 2 on Hugging Face

TradingAgents: Multi-Agents LLM Financial Trading Framework

TradingAgents proposes a multi-agent LLM framework with specialized trading roles and collaborative dynamics, outperforming baselines on cumulative returns, Sharpe ratio, and drawdown.

Xiao, Yijia, Edward W. Sun, Luo, Di, Wei Wang

Published Dec 28, 2024 · 6 citations · ▲ 150 on Hugging Face · Code ★ 110,044

Does Inference-Time Reasoning Really Improve Video Understanding?

Zhuohao Yu, Yu Zhou, Zhecan Wang, Rui Sun and 2 more

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

RAHF: Reward-Amplified Human Feedback for Closed-Loop Policy Fine-Tuning

Haoyuan Cai, Seth Zhao, Jason Zhang, Bolei Zhou

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

L$^2$EAP: Supercharging LLMs for Formal Mathematics with Agentic Frameworks

Po-Nien Kung, Linfeng Song, Dawsen Hwang, Jinsung Yoon and 9 more

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

Efficient and Accurate Zero Shot Generation of Symmetric Protein Complexes

Rory Gao, Yuanzhou Chen, Prajit Rajkumar, Wei Wang

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

Towards Scalable Egocentric HOI for Humanoids: Benchmarking Whole-Body Dexterous Interaction with Tactile Prediction

Zhenyu Wei, Haoyang Luo, Chixuan Zhang, Guo Chen and 4 more

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

Diagnosing Math-Reasoning Failure Structure with Milestone Oracles

Zhuohan Wang, Haoran Ma, Tianyu Wu, Yuanlin Duan and 2 more

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

Youdunit: Single-Call Counterfactual Necessity in Multi-Agent LLM Systems

Marissa Li, Stephanie Gao, Kenny Guo, Xingjian Li and 2 more

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

Can We Trust Item Response Theory for AI Evaluation?

Han Jiang, Sunbeom Kwon, Jinwen Luo, Ziang Xiao and 1 more

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

MetaPI: Constructing Prompt Injection Benchmarks from Any Agent Benchmarks

Peiran Wang, Chong Xiang, Wenjie Qu, Ying Li and 3 more

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

From Weeks to Hours: Fast and Principled SFT Curation for LLM

Hongyi Henry Jin, Wenhan Yang, Meysam Ghaffari, Carlos Morato and 1 more

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

TabWorld: A World-Modeling Foundation Model for Tabular Generation

Xiaofeng Lin, Chunhe Wang, Tung Sum Thomas Kwok, Guang Cheng

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

SCULPT: Advancing Masked Discrete Diffusion for High-Resolution Image Synthesis.

Shufan Li, Greg Heinrich, Hanrong Ye, Yonggan Fu and 3 more

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

PM1: A Multimodal Foundation Model for Genomes, Phenotypes, and Images at Biobank Scale

Christophe Thomassin, Marçal Comajoan Cara, Margarita Geleta, David Bonet and 3 more

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

Demystifying Numerical Errors in LLM Inference: Achieving Reproducible Inference for Mission-Critical Tasks with HEAL

Zhenting Zhu, Lucas Thai, Shan Yu, Yicheng Liu and 4 more

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

Low-Rank Hierarchical Merging for Efficient Long-to-Short Reasoning

Zeqiu Yu, Xuesheng Zhang, Wenxiao Zhao, Shixiao Wang

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

Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space

Hengli Li, Chenxi Li, Tong Wu, Xuekai Zhu and 7 more

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

Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist Rewards

Yuanhao Ban, Tong Xie, Sohyun An, Yunqi Hong and 5 more

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

TriSearch: Learning to Optimize Triangulations via Bistellar Flips

TriSearch uses reinforcement learning and circuit-based flip representations to optimize triangulations across dimensions, discovering more Calabi-Yau triangulations than existing samplers.

Yiran Wang, Guido Montufar

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

Test-Time Defense Against Adversarial Attacks via Stochastic Resonance of Latent Ensembles

A training-free test-time defense uses stochastic resonance of latent ensembles via input translations to recover up to 68.1% of adversarial accuracy loss on classification and dense prediction tasks.

DONG LAO, Yuxiang Zhang, Haniyeh E Oskouie, Yangchao Wu and 2 more

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

Token Inflation: How Dishonest Providers Can Overcharge for Large Language Model Usage

Per-token LLM billing is unauditable because providers control the evidence, enabling hidden reasoning inflation up to 1,469% and tokenization-based over-reporting of 50.85% without detection.

Shahinul Hoque, Jinghuai Zhang, Jinyuan Stella Sun, Fnu Suya

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

PhotoFlow: Agentic 3D Virtual Photography Missions

PhotoFlow uses a Director-Reviewer-Reflector agent for closed-loop camera search to generate language-conditioned virtual photographs in arbitrary 3D scenes, outperforming baselines on quality, alignment, and success rate.

Jiarui Guo, Haojia Wei, Yiming Zhang, Yifei Liu and 4 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 24 on Hugging Face · Code ★ 43

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

AMUSE: Anytime Muon with Stable Gradient Evaluation

AMUSE integrates Muon's rapid bulk progress with Schedule-Free averaging via time-varying interpolation to suppress oscillations, requiring no learning rate schedules and improving training efficiency across vision and LLM tasks.

Jueun Kim, Baekrok Shin, Jihun Yun, Beomhan Baek and 2 more

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

From Table to Cell: Attention for Better Reasoning with TABALIGN

TABALIGN improves multi-step table reasoning by pairing diffusion planners generating binary cell masks with attention verifiers, raising accuracy 15.76 points and accelerating execution 44.64%.

Tung Sum Thomas Kwok, Zeyong Zhang, Xinyu Wang, Chunhe Wang and 5 more

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

Asymmetric Phase Coding Audio Watermarking

Asymmetric Phase Coding embeds Ed25519 signatures into audio via phase-bin QIM for blind, training-free cryptographic verification resilient to cropping, compression, and resampling.

Guang Yang, Fengchen Liu, Amir Ghasemian, Zhong Wang and 2 more

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

Inertia-1: An Open Exploration of Wearable Motion Foundation Models

Inertia-1 explores wearable motion foundation models via 18.2M hours of accelerometer data, yielding state-of-the-art recipes and open design principles for diverse sensing tasks.

Zongzhe Xu, Aakarsh Anand, Sarah Jiang, Chuntung Zhuang and 3 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face · Code ★ 35

Beyond What Seems Necessary: Hidden Gains from Scaling Training-Time Reasoning Length under Outcome Supervision

Under outcome-only supervision, scaling training-time reasoning length improves OOD performance after ID saturation via stronger inductive biases and reduced shortcut reliance.

Yihao Xue, Allan Zhang, Jianhao Huang, Amit Sahai and 1 more

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

Subliminal Transfer of Unsafe Behaviors in AI Agent Distillation

Distillation of agent trajectories transfers unsafe behavioral biases subliminally despite keyword filtering, with deletion bias reaching 100% and chmod preference 30-55%.

Jacob Dang, Brian Y Xie, Omar G. Younis

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

Attention Sinks and Outliers in Attention Residuals

OASIS stabilizes dual-normalized attention-residual architectures via null routing and token-to-depth null coupling, reducing activation outliers by 81.75% and improving low-bit quantized reasoning by 42.11%.

Haozheng Luo, Haoran Dai, Shaoyang Zhang, Xi Chen and 9 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face · Code ★ 3

Trimming the Long-Tail of Visual World Modeling Evaluation

Tailor-Bench evaluates visual world models on rare physical interactions via regular, unconventional, and impossible scenarios, revealing long-tail performance gaps and superficial visual-pattern reliance.

Bingxuan Li, Yining Hong, Cheng Qian, Hyeonjeong Ha and 5 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 40 on Hugging Face · Code ★ 1

Fast Rates for Offline Contextual Bandits with Forward-KL Regularization under Single-Policy Concentrability

Forward-KL-regularized offline contextual bandits achieve epsilon^{-1} sample complexity under single-policy concentrability via pessimism, with matching lower bounds showing slow rates at weak regularization.

Qingyue Zhao, Kaixuan Ji, Heyang Zhao, Quanquan Gu

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

Learning Weakly Communicating Average-Reward CMDPs: Strong Duality and Improved Regret

Strong duality holds for weakly communicating average-reward CMDPs, yielding a primal-dual algorithm with O(T^{2/3}) regret and constraint violations.

Kihyun Yu, Beomhan Baek, Dabeen Lee

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

Stress-Testing Neural Network Verifiers with Provably Robust Instances

A framework generates provably robust verification instances with known ground-truth labels and exposes numeric errors and bugs in state-of-the-art neural network verifiers.

David Troxell, Yulia Alexandr, Sofia Hunt, Stephanie Lei and 1 more

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