On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics

In controlled strong-to-weak distillation, rollout policy is less central than token-level KL direction and learning rate, though on-policy data can improve generalization on harder reasoning tasks.

Julianna Piskorz, Antonin Berthon, Mihaela van der Schaar

Published Sep 28, 2026 · 0 citations · ▲ 194 on Hugging Face

Open-Ended Scientific Discovery and the Social Dynamics of Evolving Agent Networks

Tennison Liu, Silas Ruhrberg Estévez, Rob Davis, Ryan M Sheridan and 2 more

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

Express Language Modeling

Albert Gong, Annabelle M Carrell, Raaz Dwivedi, Lester Mackey

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

PDE-SSM: A Spectral State Space Approach to Spatial Mixing in Diffusion Transformers

Eshed Gal, Moshe Eliasof, Eldad Haber

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

Auditing AI peer reviewers: a dose-response and false-positive benchmark on real scientific papers

Íñigo Zubeldia, Boris Bolliet, Francisco Villaescusa, Pablo Villanueva-Domingo and 1 more

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

Bonobo: Efficient Library-Scale Generation for De Novo Antibody Design

Sebastian Ober, Nick Bhattacharya, Phillip M Maffettone, Calvin McCarter and 1 more

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

Contour Monte Carlo: Sampling via Energy Level Sets

Varun Jain, Hong Ge

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

Towards Closing the Autoregressive Gap in Language Modeling via Entropy-Gated Continuous Bitstream Diffusion

Georgios Batzolis, Mark Girolami, Luca Ambrogioni

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

One-Shot Private Confidence Regions via Resampling

Po-Ling Loh, Debepsita Mukherjee, Shourya Pandey, Purnamrita Sarkar

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

Forgetting to Improve: Principled Data Removal in Active Learning

Manuel Wendl, Erik Englesson, Andreas Krause, Carl Henrik Ek

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

A Subgoal-driven RL Framework for Improving Long-Horizon Web Agents

Taiyi Wang, Sian Gooding, Florian Hartmann, Oriana Riva and 1 more

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

PathNavigate: A Training-Free Pathology Agent with Surprise-Guided Scan and Shared Slide Memory for Whole-Slide VQA

Chunze Yang, Qidong Liu, Wenjie Zhao, Yue Tang and 9 more

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

LoRAtorio: An intrinsic approach to LoRA Skill Composition

Niki Foteinopoulou, Ignas Budvytis, Stephan Liwicki

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

How Language Models Compress and Compare: Understanding Selection with Token Covariance Maps

Moshe Eliasof

Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

Robust Amortized Simulation-Based Inference via Learned Error Models

Matthew O'Callaghan, Kaisey Mandel, Gerard Gilmore

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

Anchoring LLM-based Chest X-ray Report Generation via Diffusion Language Planning

Jiechao Gao, Chang Liu, Yuandong Pan, Ying Liu and 1 more

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

Learning Discrete Riemannian Metrics for Physical Fields with Cochain-Frame Equivariance

Dongzhe Zheng, Christine Allen-Blanchette

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

Position: AI Development Should Prioritize Cognitive Security

Batu El, Shiye Su, Aneesh Pappu, Peggy Yin and 5 more

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

Symmetric Interventions for Eliciting Model Intent

David Vella Zarb, Rustem Turtayev, Taywon Min, Jinghua Ou and 1 more

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

The VLM as Sensor: Bayesian Active Search for Long Video Understanding

Chong Tang, Sannara EK, Dirk Koch, Robert Mullins and 2 more

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

Breaking Curse of Dimensionality for Mutual Information Estimation with Vine Copulas

Sigurd Holmsen, Berit Øksnes, Ingrid Hobæk Haff, Sylvia Richardson and 1 more

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

Personalized Safety in Federated Fine-Tuning of Large Language Models

Tianzhe Xiao, Gaozhuo Liu, Yichen Li, Haozhao Wang and 3 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

Compatible Likelihoods for Flow Matching on Manifolds

Lucas Ng, Georgios Batzolis, Mark Girolami

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

PMO-Dock: Benchmarking Docking, Specificity, and Generalization in Molecular Optimization

Gor Simonyan, Tatevik Abrahamyan, Narek Abelyan, Tigran Fahradyan and 1 more

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

Trajectory-Consistent Diffusion Policies for Offline Reinforcement Learning

Yichao Fu, Shangde Gao, Zhuoling Li, Wen Wang and 2 more

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

Reward Budgeting Reduces Premature Convergence in Reinforcement Learning for LLM Reasoning

Mengni Jia, Mengyu Zhou, xiaoxi jiang, Guanjun Jiang

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

Scaling Causal Reasoning with Increasingly Complex Causal Simulators

Nicolás Astorga, Anita Kriz, Mihaela van der Schaar

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

We Need to Improve Benchmarks in AI for Mathematics

Simon Frieder, Jonas Bayer, Shi Zhuo Looi, Jacob Loader and 14 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

Large Language Model Failures from Hallucination to Homogenization Are Different Facets of Miscalibration

Tiancheng Hu, Caiqi Zhang, Dirk Hovy, Nigel Collier

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

Understanding Multi-View Transformers

Julien Gaubil, Michal Stary, Louis Martinez, Andreas Geiger and 3 more

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

Module-Aware Optimization for Graph Neural Networks

Guy Hadad, Haggai Roitman, Moshe Eliasof, Bracha Shapira

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

A Matter of Interest: Understanding Interestingness Judgments of Math Problems in Humans and Language Models

Shubhra Mishra, Yuka Machino, Gabriel Poesia, Albert Q. Jiang and 8 more

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

PRISM: Phenotype-Resolved Inference in Single-Cell Mixed Models via Latent Disease States and Contextualized Differential Expression

Andrea Rubbi, Lama Salem, Caleb Ellington, Pietro Lió and 3 more

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

Unified Generative-Predictive Modeling for 4D Scene Understanding

Amani Kiruga, Zhiyi Li, Ruojin Cai, Hansen Lillemark and 3 more

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

AIRA 2: Overcoming Bottlenecks in AI Research Agents

Karen Hambardzumyan, Nicolas Baldwin, Edan Toledo, RISHI HAZRA and 21 more

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

KINDER: Kernel-based Independence for Fair Representation Learning via Prototype-space Erasure

Abtin Mogharabin, Jiaee Cheong, Alp Toykan Kaplan, Sinan Kalkan

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

Generalized Intention Modeling in Multi-Agent Reinforcement Learning

A task-adaptive framework learns a performance-driven mixture of opponent intent representations to improve multi-agent reinforcement learning across diverse tasks.

Mateusz Odrowaz-Sypniewski, Jasmine Bayrooti, Ajay Shankar, Amanda Prorok

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

Generating in the Limit with Infinitely Many Hallucinations

Language generation in the limit is recast as recall-precision trade-offs, showing that allowing infinitely many vanishing-frequency hallucinations can strictly increase recall when adversaries withhold target portions.

Irene Strauss, Alexandra Butoi, Ryan Cotterell

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

HESTIA: A Hessian-Guided Differentiable Quantization-Aware Training Framework for Extremely Low-Bit LLMs

HESTIA replaces hard quantization with Hessian-guided temperature-controlled soft relaxation for low-bit LLM training, improving 1.58-bit Llama-3.2 1B and 3B zero-shot accuracy by 5.39% and 4.34%.

Guoan Wang, Feiyu Wang, Zongwei Lv, Yikun Zong and 2 more

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

Beyond Worst-Case Coreset Bounds for $k$-Clustering via Determinantal Sampling

Determinantal sampling builds smaller k-clustering coresets with sub-quadratic ε dependence under mild data assumptions, breaking worst-case bounds.

Diptarka Chakraborty, Satyaki Mukherjee, Gaurav Vallabhdas Revankar, Hoang Son Tran

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

Preconditioned Flow Matching

Ill-conditioned intermediate covariances make flow matching regress low-variance directions slowly; preconditioning into isotropic space improves optimization and generation quality.

Shadab Ahamed, Eshed Gal, Md Shahriar Rahim Siddiqui, Simon Ghyselincks and 2 more

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

AI GAMESTORE: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games

AI GameStore proposes evaluating general intelligence via scalable synthesis of human games, finding frontier vision-language models score under 10% of human averages on most generated games.

Lance Ying, Ryan Truong, Prafull Sharma, Kaiya Zhao and 8 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 9 on Hugging Face

Exploring Starts Are Not Enough: Counterexamples and a Fix for Monte Carlo Exploring Starts

Tabular Monte Carlo Exploring Starts can converge to suboptimal policies, but state-specific inverse-frequency learning-rate scaling restores convergence to optimality.

Octave Oliviers, Glenn Vinnicombe

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

MindAlign: Bridging EEG, Vision, and Language for Zero-Shot Visual Decoding

MindAlign aligns EEG, vision, and language via tri-modal contrastive learning to achieve 54.1% zero-shot visual decoding accuracy on Things-EEG2.

Zexuan Chen, Sichao Liu, Runhao Lu, Huichao Qi and 3 more

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

TimeTok: Granularity-Controllable Time-Series Generation via Hierarchical Tokenization

TimeTok introduces hierarchical tokenization for multiscale time-series generation with explicit granularity control from coarse inputs, achieving state-of-the-art results and cross-dataset transfer.

Seokhyun Lee, Jaeho Kim, Changjun Oh, Mihaela van der Schaar and 1 more

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

Tabular Foundation Model for Generative Modelling

TabFORGE introduces a tabular generative foundation model using causality-aware representations and two-stage diffusion-decoder training to generate high-fidelity synthetic data.

Xiangjian Jiang, Mingxuan Liu, Nikola Simidjievski, Tassilo Klein and 1 more

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

Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces

OPT* introduces optimization-style tasks with expanding search spaces and feasibility checkers to evaluate LLM step-by-step reasoning, showing that training on it improves optimization-like reasoning via online policy optimization and search-based offline RL.

Nicolás Astorga, Nabeel Seedat, Mihaela van der Schaar

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

Otter Weather: Skillful and computationally-efficient medium-range weather forecasting

Otter Weather achieves state-of-the-art skill with minimal compute, outperforming NWP and frontier AI weather models using under 3.5 A100-days.

Cristiana Diaconu, Jonas Scholz, Aliaksandra Shysheya, Stratis Markou and 3 more

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

Rethinking Rubric Generation for Improving LLM Judge and Reward Modeling for Open-ended Tasks

RRD refines rubrics via recursive decomposition and filtering to improve LLM judge accuracy and reinforcement training rewards on open-ended tasks.

William Shen, Xinchi Qiu, Chenxi Whitehouse, Lisa Alazraki and 5 more

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

Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors

Score errors decompose into visible gradient and invisible solenoidal parts, so L2 score error cannot bound distribution divergence and only gradient error matters for diffusion sampling quality.

Nail B Khelifa, Richard Turner, Ramji Venkataramanan

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

Recursively Trained Diffusion Models: Limiting Collapse Distribution and Spectral Characterization

Recursive diffusion training converges geometrically to a unique Gaussian-smoothed mixture limit via early-stopping drift, attenuating high-order spectral modes, with annealed truncation schedules asymptotically preventing collapse.

Nail B Khelifa, Richard Turner, Ramji Venkataramanan

Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

Learn from your own latents and not from tokens: A sample-complexity theory

Latent prediction learns hierarchical latent trees with samples constant in depth L, exponentially more efficient than token-level self-supervision, making explicit multi-scale stacking largely redundant.

Daniel Korchinski, Alessandro Favero, Matthieu Wyart

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 3 on Hugging Face

SLVMBench: Skill Learning from Video Memory

SLVMBench evaluates video-LLMs on learning skills from long video streams and applying them in real time, revealing substantial performance degradation.

Yudong Yang, Guangzhi Sun, Yixuan Li, Wei Li and 2 more

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

Follow the Winners: Conservative Policy Improvement with the Cross-Entropy Method for Critic-Free RFT

FTW replaces GRPO group rollouts with replay-buffer ordinal filtering for critic-free agentic RFT, matching PPO and GRPO on Sokoban and Search-R1 while trading value models or rollouts for CPU memory.

Joery A de Vries, Neil Lawrence, Zhenwen Dai

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

OpenMHC: Accelerating the Science of Wearable Foundation Models

OpenMHC releases the largest open wearable health dataset with open-source foundation models and a unified benchmark across prediction, imputation, and forecasting tasks.

Narayan Schütz, Yuze Bai, Lianggang Pan, Edgar Eggert and 15 more

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

Instruct-Particulate: Scaling Feed-Forward 3D Object Articulation with Kinematic Control

Instruct-Particulate predicts articulated 3D part segmentation and joint parameters from meshes and kinematic specifications, scaling training via vision-language labels to improve cross-category and AI-generated mesh generalization.

Ruining Li, Yuxin Yao, Matt Zhou, Chuanxia Zheng and 4 more

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

BankerToolBench: Evaluating AI Agents in End-to-End Investment Banking Workflows

BankerToolBench benchmarks AI agents on multi-hour investment banking workflows using expert rubrics, finding frontier models fail nearly half of criteria with zero client-ready outputs.

Elaine Lau, Markus Dücker, Ronak Chaudhary, Hui Wen Goh and 24 more

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

GlucoFM: A Dual-Stream Foundation Model for Continuous Glucose Monitoring

GlucoFM decomposes CGM data into dual slow and short-term streams for pretraining, improving linear-probe phenotype classification and postprandial response prediction over prior models.

Zechen Li, Keerthana Natarajan, Weizhi Zhang, Simon Lee and 10 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 8 on Hugging Face

Neural Field Thermal Tomography: A Differentiable Physics Framework for Non-Destructive Evaluation

NeFTY uses a hard-constrained differentiable implicit Euler solver and neural fields to recover 3D thermal diffusivity from surface measurements, outperforming soft PINNs and classical baselines on synthetic and real thermography.

Tao Zhong, Yixun Hu, Dongzhe Zheng, Aditya Sood and 1 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 2 on Hugging Face

Learning Energy-Based Models from Stochastic Interpolants using Spatiotemporal Differences

Spatiotemporal Noise-Contrastive Estimation learns energy-based models via joint spatiotemporal differences to avoid failure modes of spatial or temporal methods alone, matching state-of-the-art density estimation.

Hanlin Yu, RuiKang OuYang, Partha Kaushik, Arto Klami and 2 more

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

ThousandWorlds: A benchmark for climate emulation of potentially habitable exoplanets

ThousandWorlds introduces a multi-model exoplanet climate benchmark of ~1,700 GCM simulations, showing Gaussian processes outperform deep learning in low-data multi-simulator regression.

Edward Stevenson, Mei T Mak, Eric Wolf, Denis E Sergeev and 3 more

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

Get a GRIP, this will be a long TRIP: A Quantifiable Long-Range Framework for Verifying Over-squashing

Introducing verifiable axioms for long-range graph benchmarks, this work proposes TRIP/GRIP to construct provably long-range tasks with closed-form per-range error bounds and audits existing benchmarks.

Ferran Hernandez Caralt, Simon Heilig, Adrián Bazaga, Asja Fischer and 2 more

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

The Design Space of Tri-Modal Masked Diffusion Models

A tri-modal masked diffusion model pretrained from scratch on text, image-text, and audio-text data achieves strong cross-modal generation and introduces an SDE-based batch-size reparameterization.

Louis Bethune, Victor Guilherme Turrisi da Costa, Bruno Mlodozeniec, Pau Rodriguez and 20 more

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

How to make the most of your masked language model for protein engineering

Stochastic beam search samples masked protein language models via pseudo-perplexity to flexibly optimize sequences, with in vitro antibody tests showing sampling choices substantially affect engineering success.

Calvin McCarter, Nick Bhattacharya, Sebastian Ober, Hunter Elliott

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

Understanding Goal Generalisation in Sequential Reinforcement Learning

Salient features drive reinforcement learning goal generalization, early goals persist to affect later ones, and latent policy gradients predict out-of-distribution behavior accurately.

Jason Brown, Edward Young

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

SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild

SAM 3D Animal is a promptable framework using the SMAL+ model and Herd3D dataset to reconstruct multiple 3D animals from single images with keypoint and mask prompts, achieving state-of-the-art results.

Xuyi Hu, Jin Lyu, Jiuming Liu, Yebin Liu and 3 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 3 on Hugging Face · Code ★ 13

Self-Improving World Modelling with Latent Actions

SWIRL learns world models from state-only sequences via latent actions and alternating forward/inverse dynamics, improving LLM/VLM reasoning benchmarks by up to 28%.

Yifu QIU, Zheng Zhao, Waylon Li, Yftah Ziser and 3 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 32 on Hugging Face · Code ★ 20

Continually Evolving Skill Knowledge in Vision Language Action Model

Stellar VLA learns evolving skills via parameter-free continual imitation with knowledge-guided routing, achieving strong LIBERO performance with 1% replay and real-world transfer.

Yuxuan Wu, Guangming Wang, Zhiheng Yang, Tianchen Deng and 3 more

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

Conditioning Gaussian Processes on Almost Anything

Gaussian processes are recast as linear diffusion models to enable conditioning on arbitrary likelihoods, including language and physics, via ODE sampling without bespoke derivations.

Henry Moss, Lachlan Astfalck, Tom Cowperthwaite, Colin Doumont and 4 more

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

Don't Lose Focus: Activation Steering via Key-Orthogonal Projections

SKOP constrains harmful attention rerouting by preserving focus-token attention during steering, reducing utility degradation 5-7x at over 95% steering efficacy.

Haoyan Luo, Mateo Espinosa Zarlenga, Mateja Jamnik

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

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

Adaptive Coordinate Transforms for Neural Operators

ACT introduces a plug-and-play block that learns adaptive coordinate transforms for neural operators, reducing spatial misalignment and significantly improving predictive accuracy across PDE benchmarks.

Chaoyu Liu, Zhonghao Li, Gaohang Chen, Zakhar Shumaylov and 4 more

Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

Bench-MFG: A Benchmark Suite for Learning in Stationary Mean Field Games

Bench-MFG proposes a standardized discrete-time mean field game benchmark suite with a problem taxonomy, random instance generation, and evaluation guidelines for learning algorithms.

Lorenzo Magnino, Jiacheng Shen, Matthieu Geist, Olivier Pietquin and 1 more

Paris Poster Session 6, Fri, Dec 11, 2:30 PM–4:30 PM, Paris Poster Hall · Published 2026

Liars' Bench: Evaluating Lie Detectors for Language Models

Liars' Bench evaluates lie detectors across 72,863 LLM lies and finds existing techniques systematically miss certain lie types, especially when transcripts alone are insufficient.

Kieron Kretschmar, Walter Laurito, Sharan Maiya, Samuel Marks

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026 · ▲ 1 on Hugging Face

Dynamic Expert Sharing: Decoupling Memory from Parallelism in Mixture-of-Experts Diffusion LLMs

DES selects sequence-level expert coresets for diffusion MoE LLMs to cut unique activations over 55% and latency up to 38% while retaining 99% accuracy, decoupling memory from parallelism.

Hao Chen, Zhiwen Mo, Royson Lee, Qianzhou Wang and 5 more

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

Articraft: An Agentic System for Scalable Articulated 3D Asset Generation

Articraft uses LLM agents to programmatically generate validated articulated 3D assets at scale, yielding 10K assets for training and simulation.

Matt Zhou, Ruining Li, Xiaoyang Lyu, Zhaomou Song and 5 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

LOFT: Low-Rank Orthogonal Fine-Tuning via Task-Aware Support Selection

LOFT separates orthogonal PEFT subspaces from transformations to enable task-aware support selection, improving efficiency-performance trade-offs across language, vision, and reasoning tasks.

Lanxin Zhao, Bamdev Mishra, Pratik Kumar Jawanpuria, Lequan Lin and 3 more

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

Few-Step Boltzmann Generators via Scalable Likelihood Flow Maps

SCALLOP introduces a Hutchinson-free likelihood distillation objective for few-step Boltzmann generators, reducing training variance and time while achieving up to 10x inference speedup.

RuiKang OuYang, Hanlin Yu, Xinyue Ai, Yutong He and 6 more

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

Free energy Estimation on Any State Space

The paper generalizes neural-transport-accelerated free energy estimation to arbitrary state spaces, validating it across discrete, multimodal, and autoregressive settings while establishing group-theoretic identities linking time reversal and Doob's transforms.

Jiajun He, Zijing Ou, Francisco Vargas, Yingzhen Li and 3 more

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

TriAxialKV: Toward Extreme Low-Precision KV-Cache Quantization for Agentic Inference Tasks

TriAxialKV assigns triaxial tags to KV-cache tokens and uses per-tag sensitivity to allocate INT2/INT4 under fixed memory, matching BF16 accuracy with 4.5x cache compression and 30% higher throughput on agentic tasks.

Hanzhang Shen, Haoran Wu, Yiren Zhao, Robert Mullins

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

Theory of Mind and Persuasion Beyond Conversation: Assessing the Capacity of LLMs to Induce Belief States via Planning and Action

LLMs evaluated on Non-Conversational Planning ToM via object manipulation to induce belief states; GPT-5 achieved ~80% success, outperforming humans but remaining less robust, with all models better at inducing true than false beliefs.

Ben Slater, Lucy G Cheke, John Burden, Winnie Street

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