Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models

Modality-aware neuron pruning removes target knowledge from multimodal LLMs by selectively pruning modality-specific neurons to enable precise unlearning.

Zheyuan Liu, Guangyao Dou, Xiangchi Yuan, Chunhui Zhang and 2 more

Published 2025 · 4 citations

Inference and Uncertainty Quantification for Streaming $r$-PCA

Haoshu Xu, Hongzhe Li

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

Do Diffusion Models Learn to Generalize Basic Visual Skills?

Amish Sethi, Boya Zeng, Wenhao Chai, Zhuang Liu

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

Point Clustering Encoders

Evangelos Chatzipantazis, Guillem Brasó, Cristiano Saltori, Sérgio Agostinho and 3 more

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

CupOFMoCA: Coupled Objective-Guided Discrete Flows for Molecular Conjugate Assembly

Ruoxi Zhang, Ziang Li, Jiatao Gu, Pranam Chatterjee

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

Scaling Neural Motor Decoding via Decoupled Behavioral Pretraining

Divyansha ., Vinam Arora, Shivashriganesh P. Mahato, Alexandre Andre and 1 more

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

Robust and Efficient Backdoor Mitigation for ML Models via Tolerant Property Testing

Xi Chen, Anindya De, Rocco A Servedio

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

Position-aware eXplanation: A Model-Agnostic Framework for Positional Attributions

Chaehyeon Kim, Gary Weissman, Eric Wong

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

Exact Topological Compliance: Generating Persistence-Equivalent Graphs

Mattie Ji, Indradyumna Roy, Vikas Garg

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

AlloGen: Conformation-Selective Binder Design with Differential State Scoring

Hanqun Cao, Aastha Pal, Sumi Kimura, Yesol Kim and 3 more

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

What Sound Tells You About the Room

Yiduo Hao, Yiwei Tang, Jiayang Li, Zitong Lan and 3 more

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

Reliable Abstention under Adversarial Injections: Lower Bounds and New Upper Bounds

Ezra Edelman, Surbhi Goel

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

Position: The Road to Generalizable Neuro-Symbolic Learning Should be Paved with Foundation Models

Adam Stein, Aaditya Naik, Neelay Velingker, Mayur Naik and 1 more

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

Narrowing the Collaboration Gap, Probably

Mirah Shi, Marcel Hussing, Natalie Collina, Ira Globus-Harris and 2 more

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

Transfer Entropy as a Measure of Information Flow in VLMs and LLMs

Jessica E Liang, Jianbo Shi

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

Ground False: Uncovering Errors in Formal Mathematics Benchmarks

Marcus Min, One An, Xujie Si, Osbert Bastani

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

What Fits (Into Few Tokens) Doesn't Overfit: Compression and Generalization in ML Research Agents

LLM research agents find high-performance models via compressed prompts and feedback, supporting a description-length explanation for limited overfitting.

Martin Bertran, Aaron Roth, Steven Wu

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

Learning to target with network interference

Adaptive targeting under sparse network interference achieves near-optimal regret depending on structural knowledge, proving standard linear bandits are inefficient and offering practical algorithms.

Xiaomeng Wang, Hamsa Bastani, Osbert Bastani, Zhimei Ren

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

DiPhon: Diffusion on Graphons for Scalable Graph Generation

DiPhon defines graphon diffusion via a Jacobi SDE for scalable graph generation, matching first moments exactly and preserving topology across sizes without retraining.

Sergio Rozada, Yiming QIN, Manuel Madeira, Pascal Frossard and 1 more

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

Invertible Logits Transformation for Accuracy-Preserving Post-Hoc Uncertainty Calibration

InvLT calibrates uncertainties by applying a shared monotonic scalar MLP to logits, preserving predictions independent of class count and outperforming baselines on standard benchmarks.

Lening Zhao, Qipeng Zhan, Li Shen

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

Search at the Cost of Sampling: Nearly-Instant Latent Space Bayesian Optimization

Exploiting spherical latent geometry yields nearly closed-form Bayesian optimization with 100x speedups and matching performance in generative discovery pipelines.

Donney Fan, Colin Doumont, Aleksandra Kalisz, Paul Duckworth and 3 more

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

Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves

A Hilbert bundle convolutional framework defines HilbNets for infinite-dimensional manifold signals, proving discrete versions converge to continuous architectures and transfer across samplings.

Kartik Tandon, Julian J Gould, Tanishq Bhatia, Francesca Dominici and 2 more

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

Strategic Decision Support for AI Agents

Strategic decision support minimizes AI agent support usage via threshold policies controlling counterfactual missed-support error without distributional assumptions.

Shayan Kiyani, Sima Noorani, George J. Pappas, Hamed Hassani

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

ELSA3D: Elastic Semantic Anchoring for Unified 3D Understanding and Generation

ELSA3D introduces elastic semantic anchoring to unify 3D understanding and generation via scale-matched cross-modal routing, achieving state-of-the-art results with roughly half the FLOPs and latency.

Tianjiao (Joey) Yu, Xinzhuo Li, Yifan Shen, Onkar Susladkar and 3 more

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

SWE Atlas: Benchmarking Coding Agents Beyond Issue Resolution

SWE Atlas benchmarks coding agents on codebase Q&A, test writing, and refactoring, finding frontier models lead but all struggle with edge cases and engineering quality.

Mohit Raghavendra, Soham Dan, Miguel Romero Calvo, Yannis He and 11 more

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

Provable Test-Time Scaling for Beam Search in LLM Reasoning

Modified confidence-filtered beam search achieves near-linear token-level coverage scaling versus quadratic for vanilla beam search, with polynomial horizon dependence that outperforms exponentially scaling Best-of-N methods.

Qijia He, Yu Huang, Yuan Cheng, Yuxin Chen and 1 more

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

Knowledge Transfer Scaling Laws for 3D Medical Imaging

Medical imaging pretraining reveals asymmetric cross-domain scaling and power-law transfer, yielding optimized data allocations with a hub-and-island structure that improves transfer over proportional sampling by up to 58%.

Ho Hin Lee, Dongna Du, Chu Wang, Yuankai Huo and 3 more

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

The SuperActivator Mechanism: Transformers Concentrate Reliable Concept Signals in the Tail

Transformers use superactivator mechanisms to amplify concept activation gaps, concentrating reliable evidence into sparse high-activation token tails that improve concept detection F1 by up to 0.14.

Cassandra Goldberg, Chaehyeon Kim, Adam Stein, Eric Wong

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

Active Flow Expansion for Out-of-Distribution Discovery: from Theory to Molecules

Active Flow Expansion uses verifier-guided active exploration to grow a flow model's generable set, yielding theoretical guarantees and superior out-of-distribution molecule and protein design.

Riccardo De Santi, Bruce D Lee, Cristian Jensen, Kimon Protopapas and 5 more

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

pCoMole: Pareto-Constrained Molecule Editing with Discrete Flows

pCoMole uses discrete flow matching to edit biomolecular sequences toward Pareto-optimal targets while enforcing hard biochemical and manufacturability constraints. Wet-lab tests show edited eGFP variants retain fluorescence after deletions and substitutions.

Tong Chen, Maximilian Holsman, Lin Zhao, Yinuo Zhang and 1 more

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

Instance-Adaptive Online Multicalibration

An efficient online multicalibration algorithm adaptively refines a dyadic grid to interpolate between worst-case and benign sequences, achieving rates from O(T^{2/3}) down to O(sqrt(T)) and O(sqrt(JT)) with tight threshold-complexity dependence.

Zhiming Huang, Jamie Morgenstern, Aaron Roth, Claire Jie Zhang

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

Beyond Right and Wrong: Evaluating Second-order Social Reasoning in Large Language Models

LLMs overpredict social punishment relative to human judgments and align less with distant observers, revealing distorted second-order metanorm reasoning.

Sunny Rai, Jinyi Kuang, Reyhan Jamalova, Niyati Malhotra and 6 more

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

Beyond Domains: Reusing Web Skills via Transferable Interaction Patterns

SkillMigrator learns reusable web skills via transferable interaction patterns matched by layout similarity to reduce LLM actions 8-10% across WebArena and Mind2Web.

Shiqi He, Yue Cui, Feijie Wu, Xinyu Ma and 4 more

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

FlowBank: Query-Adaptive Agentic Workflows Optimization through Precompute-and-Reuse

FlowBank optimizes LLM agent workflows via a precomputed diverse portfolio selected per query, outperforming baselines by up to 14.92% at competitive cost.

Lingzhi Yuan, Chenghao Deng, Fangxu Yu, Souradip Chakraborty and 2 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 2 on Hugging Face

Robust Policy Optimization to Prevent Catastrophic Forgetting

FRPO robustifies RLHF via a max-min objective that maximizes reward across downstream-adaptation neighborhoods, substantially reducing catastrophic forgetting while preserving task performance.

Mahdi Sabbaghi, George J. Pappas, Adel Javanmard, Hamed Hassani

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · Code ★ 6

Self Driving Datasets: From 20 Million Papers to Nuanced Biomedical Knowledge at Scale

PubMed is autonomously converted into structured biomedical datasets larger, more nuanced, and more accurate than manual repositories via ontology tagging, hybrid retrieval, and a multi-agent extraction system.

Haydn Jones, Yimeng Zeng, Alden Rose, Yifei Li and 10 more

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

LogSTOP: Temporal Scores over Prediction Sequences for Matching and Retrieval

LogSTOP computes Linear Temporal Logic temporal property scores over noisy local prediction sequences, outperforming language models and retrieval baselines by at least 16% on matching and retrieval.

Avishree Khare, Hideki Okamoto, Bardh Hoxha, Georgios Fainekos and 1 more

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

Every Sample Counts: Supervised Fine-Tuning of Language Models with Pointwise Constraints

A framework enforces per-sample constraints during language-model fine-tuning via learnable relaxations and augmented Lagrangians, reducing tail violations while preserving performance.

Ignacio Hounie, Ignacio Boero, Alejandro Ribeiro

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

AdaptNC: Adaptive Nonconformity Scores for Conformal Prediction under Distribution Shift

AdaptNC jointly adapts nonconformity scores and conformal thresholds online to reduce prediction volumes under distribution shift while maintaining coverage.

Renukanandan Tumu, Aditya Singh, Rahul Mangharam

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

When Reasoning Meets Its Laws

This paper proposes Laws of Reasoning (LoRe), a framework formalizing reasoning compute and accuracy laws, plus LoRe-Bench showing models lack compositionality; enforcing compute-law compositionality via finetuning improves reasoning performance.

Junyu Zhang, Yifan Sun, Tianang Leng, Jingyan Shen and 3 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 62 on Hugging Face · Code ★ 38