Isolater - Feed

Ax Zixuan Huang, Xin Xia, Yuxi Ren, Jianbin Zheng, Xuanda Wang, Zhixia Zhang, Hongyan Xie, Songshi Liang, Zehao Chen, Xuefeng Xiao, Fuzhen Zhuang, Jianxin Li, Yikun Ban, Deqing Wang 3/2/2026

Does Your Reasoning Model Implicitly Know When to Stop Thinking?

Study on whether Large Reasoning Models know when to stop thinking, addressing redundancy in long chains-of-thought.

Ax Zewei Yu, Lirong Gao, Yuke Zhu, Bo Zheng, Junbo Zhao, Sheng Guo, Haobo Wang 3/2/2026

Stop Unnecessary Reflection: Training LRMs for Efficient Reasoning with Adaptive Reflection and Length Coordinated Penalty

Training method for Large Reasoning Models using adaptive reflection and length penalties to reduce unnecessary token consumption.

Ax Haibo Tong, Feifei Zhao, Linghao Feng, Ruoyu Wu, Ruolin Chen, Lu Jia, Zhou Zhao, Jindong Li, Tenglong Li, Erliang Lin, Shuai Yang, Enmeng Lu, Yinqian Sun, Qian Zhang, Zizhe Ruan, Jinyu Fan, Zeyang Yue, Ping Wu, Huangrui Li, Chengyi Sun, Yi Zeng 3/2/2026

ForesightSafety Bench: A Frontier Risk Evaluation and Governance Framework towards Safe AI

ForesightSafety Bench evaluates frontier risks in autonomous AI with unpredictable and difficult-to-control behaviors.

Ax Seoyoung Lee, Seobin Yoon, Seongbeen Lee, Yoojung Chun, Dayoung Park, Doyeon Kim, Joo Yong Sim 3/2/2026

IntentCUA: Learning Intent-level Representations for Skill Abstraction and Multi-Agent Planning in Computer-Use Agents

IntentCUA framework for computer-use agents with intent-aligned planning and multi-agent coordination over long horizons.

Ax Tao Zhe, Haoyu Wang, Bo Luo, Min Wu, Wei Fan, Xiao Luo, Zijun Yao, Haifeng Chen, Dongjie Wang 3/2/2026

Robust and Efficient Tool Orchestration via Layered Execution Structures with Reflective Correction

Layered execution structures for tool orchestration in agentic systems with reflective error correction mechanisms.

Ax Tony Feng, Junehyuk Jung, Sang-hyun Kim, Carlo Pagano, Sergei Gukov, Chiang-Chiang Tsai, David Woodruff, Adel Javanmard, Aryan Mokhtari, Dawsen Hwang, Yuri Chervonyi, Jonathan N. Lee, Garrett Bingham, Trieu H. Trinh, Vahab Mirrokni, Quoc V. Le, Thang Luong 3/2/2026

Aletheia tackles FirstProof autonomously

Aletheia AI agent solved 6/10 FirstProof mathematics challenges autonomously using Gemini 3 Deep Think reasoning.

Ax Willem Schooltink, Fabio Massimo Zennaro 3/2/2026

Multi-Level Causal Embeddings

Framework for causal embeddings enabling multiple detailed models to map into sub-systems of coarser causal models.

Ax Yongjun Zhang 3/2/2026

Vibe Researching as Wolf Coming: Can AI Agents with Skills Replace or Augment Social Scientists?

Examines AI agents with persistent state, tool access, and skills for autonomous execution of social science research pipelines.

Ax Joseph Tso, Preston Schmittou, Quan Huynh, Jibran Hutchins 3/2/2026

ConstraintBench: Benchmarking LLM Constraint Reasoning on Direct Optimization

ConstraintBench evaluates whether LLMs can directly solve constrained optimization problems without solver access.

Ax Xun Huang, Simeng Qin, Xiaoshuang Jia, Ranjie Duan, Huanqian Yan, Zhitao Zeng, Fei Yang, Yang Liu, Xiaojun Jia 3/2/2026

Obscure but Effective: Classical Chinese Jailbreak Prompt Optimization via Bio-Inspired Search

Study on LLM vulnerability to jailbreak attacks using classical Chinese prompts to bypass safety constraints.

Ax Martin Riedmiller, Andrea Gesmundo, Tim Hertweck, Roland Hafner 3/2/2026

Less is more -- the Dispatcher/ Executor principle for multi-task Reinforcement Learning

Dispatcher/Executor principle for multi-task reinforcement learning using abstraction to improve generalization across tasks.

Ax Xiao Wang, Yuehang Li, Fuling Wang, Shiao Wang, Chuanfu Li, Bo Jiang 3/2/2026

R2GenCSR: Mining Contextual and Residual Information for LLMs-based Radiology Report Generation

R2GenCSR uses LLMs with visual feature extraction from X-ray images for automated radiology report generation.

Ax Lei You, Yijun Bian, Lele Cao 3/2/2026

Joint Distribution-Informed Shapley Values for Sparse Counterfactual Explanations

Research framework for sparse counterfactual explanations using optimal transport and Shapley values for model interpretability.

Ax Mikhail Pautov, Danil Ivanov, Andrey V. Galichin, Oleg Rogov, Ivan Oseledets 3/2/2026

Spread them Apart: Towards Robust Watermarking of Generated Content

Research on robust watermarking techniques for distinguishing generated from real content in generative models.

Ax Ankur Sinha, Chaitanya Agarwal, Pekka Malo 3/2/2026

FinBloom: Knowledge Grounding Large Language Model with Real-time Financial Data

Research paper on grounding LLMs with real-time financial data for knowledge-aware financial agent applications.

Ax Yan Li, Zhenyu Zhang, Zhengang Wang, Pengfei Chen, Pengfei Zheng 3/2/2026

Semantic Parallelism: Redefining Efficient MoE Inference via Model-Data Co-Scheduling

Semantic parallelism technique for efficient MoE LLM inference via model-data co-scheduling reducing communication bottlenecks.

Ax Noam Razin, Zixuan Wang, Hubert Strauss, Stanley Wei, Jason D. Lee, Sanjeev Arora 3/2/2026

What Makes a Reward Model a Good Teacher? An Optimization Perspective

Optimization perspective on reward model quality in RLHF showing accuracy alone doesn't capture effective teacher properties.

Ax Jianing Huang, Kaixuan Zhang, Youjia Wu, Ze Cheng 3/2/2026

Operator Learning with Domain Decomposition for Geometry Generalization in PDE Solving

Domain decomposition approach for neural operators to improve geometry generalization and transferability in PDE solving.

Ax Lingyan Bao, Sinwoong Yun, Jemin Lee, Tony Q. S. Quek 3/2/2026

LLM-hRIC: LLM-empowered Hierarchical RAN Intelligent Control for O-RAN

LLM-empowered hierarchical RIC controller for O-RAN addressing cooperation, computational demands, and domain-specific adaptation.

Ax Chaitali Bhattacharyya, Hyunsei Lee, Junyoung Lee, Shinhyoung Jang, Il hong Suh, Yeseong Kim 3/2/2026

FineScope : SAE-guided Data Selection Enables Domain Specific LLM Pruning and Finetuning

FineScope framework using SAE-guided data selection for domain-specific LLM pruning and finetuning with maintained performance.

Ax Rui Liu, Rui Xie, Zijun Yao, Yanjie Fu, Dongjie Wang 3/2/2026

Continuous Optimization for Feature Selection with Permutation-Invariant Embedding and Policy-Guided Search

Feature selection method using permutation-invariant embeddings and policy-guided search for complex feature interactions.

Ax Patara Trirat, Wonyong Jeong, Sung Ju Hwang 3/2/2026

Multi-View Encoders for Performance Prediction in LLM-Based Agentic Workflows

Agentic Predictor using multi-view encoders for performance prediction in LLM-based agentic workflows without exhaustive evaluation.

Ax Th\'eo Vincent, Yogesh Tripathi, Tim Faust, Abdullah Akg\"ul, Yaniv Oren, Melih Kandemir, Jan Peters, Carlo D'Eramo 3/2/2026

Bridging the Performance Gap Between Target-Free and Target-Based Reinforcement Learning

Method bridging target-free and target-based deep reinforcement learning to reduce memory requirements and improve update propagation.

Ax Yeong-Joon Ju, Seong-Whan Lee 3/2/2026

From Generator to Embedder: Harnessing Innate Abilities of Multimodal LLMs via Building Zero-Shot Discriminative Embedding Model

Framework converting generative multimodal LLMs into zero-shot discriminative embedding models without extensive pre-training.

Ax Zhuoran Li, Xun Wang, Hai Zhong, Qingxin Xia, Lihua Zhang, Longbo Huang 3/2/2026

OM2P: Offline Multi-Agent Mean-Flow Policy

OM2P offline multi-agent reinforcement learning using flow-based generative models with improved sampling efficiency.

Ax Arjun Ashok, Andrew Robert Williams, Vincent Zhihao Zheng, Irina Rish, Nicolas Chapados, \'Etienne Marcotte, Valentina Zantedeschi, Alexandre Drouin 3/2/2026

Beyond Na\"ive Prompting: Strategies for Improved Context-aided Forecasting with LLMs

Framework for improving LLM context-aided forecasting with diagnostic tools and reduced computational costs for practical deployment.

Ax Jiarui Yang, Bin Zhu, Jingjing Chen, Yu-Gang Jiang 3/2/2026

Actor-Critic for Continuous Action Chunks: A Reinforcement Learning Framework for Long-Horizon Robotic Manipulation with Sparse Reward

AC3 reinforcement learning framework for long-horizon robotic manipulation using continuous action chunking with sparse rewards.

Ax Ron Solomon, Yarin Yerushalmi Levi, Lior Vaknin, Eran Aizikovich, Amit Baras, Etai Ohana, Amit Giloni, Shamik Bose, Chiara Picardi, Yuval Elovici, Asaf Shabtai 3/2/2026

LumiMAS: A Comprehensive Framework for Real-Time Monitoring and Enhanced Observability in Multi-Agent Systems

LumiMAS framework for real-time monitoring and observability of multi-agent systems with LLMs, addressing system-wide failure detection.

Ax Alexander Steen 3/2/2026

A Reduction of Input/Output Logics to SAT

SAT reduction approach for automating input/output logics, a family of deontic logics for reasoning over norms and obligations.

Ax Jungsuk Oh, Jay-Yoon Lee 3/2/2026

Latent Self-Consistency for Reliable Majority-Set Selection in Short- and Long-Answer Reasoning

Latent Self-Consistency: method for reliable majority voting in LLM outputs handling both short and long-form reasoning tasks consistently.

Ax Maolin Sun, Yibiao Yang, Yuming Zhou 3/2/2026

Once4All: Skeleton-Guided SMT Solver Fuzzing with LLM-Synthesized Generators

Once4All: LLM-synthesized test generator for SMT solver fuzzing using skeleton guidance to uncover bugs in evolving solver versions.

Ax Hao Tan, Jun Lan, Zichang Tan, Ajian Liu, Chuanbiao Song, Senyuan Shi, Huijia Zhu, Weiqiang Wang, Jun Wan, Zhen Lei 3/2/2026

Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning

Veritas: pattern-aware deepfake detection system with HydraFake dataset bridging gap between academic benchmarks and industrial deployment.

Ax Ziyun Zeng, David Junhao Zhang, Wei Li, Mike Zheng Shou 3/2/2026

Draw-In-Mind: Rebalancing Designer-Painter Roles in Unified Multimodal Models Benefits Image Editing

Draw-In-Mind: multimodal model rebalancing designer and painter roles to improve precision in image editing tasks.

Ax Mengqi Wang, Zhan Liu, Zengrui Jin, Guangzhi Sun, Chao Zhang, Philip C. Woodland 3/2/2026

Audio-Conditioned Diffusion LLMs for ASR and Deliberation Processing

LLaDA diffusion-based large language model applied to automatic speech recognition with deliberation-based post-processing for Whisper transcripts.

Ax Zhengkang Guan, Kun Kuang 3/2/2026

Efficient Ensemble Conditional Independence Test Framework for Causal Discovery

E-CIT: plug-and-play ensemble framework for conditional independence testing to reduce computational bottlenecks in constraint-based causal discovery.

Ax Fan Wang, Zhiyuan Chen, Yuxuan Zhong, Sunjian Zheng, Pengtao Shao, Bo Yu, Shaoshan Liu, Jianan Wang, Ning Ding, Yang Cao, Yu Kang 3/2/2026

Context and Diversity Matter: The Emergence of In-Context Learning in World Models

Investigation of in-context learning emergence in world models for environmental dynamics prediction beyond static zero-shot performance.

Ax Lute Lillo, Nick Cheney 3/2/2026

Activation Function Design Sustains Plasticity in Continual Learning

Study of activation function design's role in preventing plasticity loss during continual learning, beyond catastrophic forgetting.

Ax Junming Yang, Ning Xu, Biao Liu, Shiqi Qiao, Xin Geng 3/2/2026

Alignment through Meta-Weighted Online Sampling: Bridging the Gap between Data Generation and Preference Optimization

Meta-weighted online sampling approach for aligning LLMs by reducing distribution mismatch between offline preference data and evolving model policy.

Ax Changsheng Zhao, Ernie Chang, Zechun Liu, Chia-Jung Chang, Wei Wen, Chen Lai, Sheng Cao, Yuandong Tian, Raghuraman Krishnamoorthi, Yangyang Shi, Vikas Chandra 3/2/2026

MobileLLM-R1: Exploring the Limits of Sub-Billion Language Model Reasoners with Open Training Recipes

MobileLLM-R1: sub-billion parameter language models with chain-of-thought reasoning and open training recipes, challenging assumptions about model size requirements.

Ax Shuofei Qiao, Yanqiu Zhao, Zhisong Qiu, Xiaobin Wang, Jintian Zhang, Zhao Bin, Ningyu Zhang, Yong Jiang, Pengjun Xie, Fei Huang, Huajun Chen 3/2/2026

Scaling Generalist Data-Analytic Agents

DataMind: scalable data-analytic agent system with open-source training recipes for multi-step reasoning over diverse-format, large-scale data files.

Ax Guancheng Chen, Sheng Yang, Tong Zhan, Jian Wang 3/2/2026

BEV-VLM: Trajectory Planning via Unified BEV Abstraction

BEV-VLM: trajectory planning approach using vision-language models with bird's-eye-view representations from fused camera and LiDAR data.

Ax Chi Zhang, Kaiwen Zheng, Zehua Chen, Jun Zhu 3/2/2026

VoiceBridge: General Speech Restoration with One-step Latent Bridge Models

VoiceBridge: one-step latent bridge model for general speech restoration from diverse distortions with energy-preserving VAE design.

Ax Sheng Yang, Tong Zhan, Guancheng Chen, Yanfeng Lu, Jian Wang 3/2/2026

Less is More: Lean yet Powerful Vision-Language Model for Autonomous Driving

Max-V1: vision-language model framework for autonomous driving that formulates trajectory planning as next waypoint prediction via language.

Ax Marcel Wien\"obst, Leonard Henckel, Sebastian Weichwald 3/2/2026

Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning

FLOP: score-based causal discovery algorithm for linear models using fast parent selection and Cholesky-based updates to find optimal causal graphs.

Ax Haining Pan, James V. Roggeveen, Erez Berg, Juan Carrasquilla, Debanjan Chowdhury, Surya Ganguli, Federico Ghimenti, Juraj Hasik, Henry Hunt, Hong-Chen Jiang, Mason Kamb, Ying-Jer Kao, Ehsan Khatami, Michael J. Lawler, Di Luo, Titus Neupert, Xiaoliang Qi, Michael P. Brenner, Eun-Ah Kim 3/2/2026