Wiredframe Radar logo

Category intelligence

Research Briefing — February 3, 2026

1392 current items analyzed and ranked.

Executive synthesis

Research Summary

Today's research features potentially transformative efficiency advances and critical safety findings. A symmetry-aware Taylor approximation claims to achieve constant-cost self-attention per token—if validated, a fundamental breakthrough. Meta introduces Fault Tolerant HSDP enabling training on 100K+ GPUs with graceful failure recovery.

Theoretical advances include polylog(1/δ) sampling complexity for diffusion models (exponential improvement), formal proofs that transformers learn factored representations in orthogonal subspaces, and a relative-budget theory explaining when RLVR succeeds. Grad2Reward extracts dense process rewards directly from LLM judge gradients, addressing reward sparsity in long-form reasoning.

Key Themes

AI Safety & Alignment · 49Efficiency & Scaling · 10LLM Reasoning & Interpretability · 8Vision-Language-Action Models · 18AI Safety & Security · 19LLM Training & RL · 8RLHF & Alignment · 7LLM Training and Alignment · 8AI Safety & Fairness · 8Reasoning & Robustness · 6

Primary evidence

Top Ranked Signals

Research arXiv (Artificial Intelligence) Feb 3

Self-Attention at Constant Cost per Token via Symmetry-Aware Taylor Approximation

By Franz A. Heinsen, Leo Kozachkov

92 score
AI Analysis

Shows self-attention is efficiently computable to arbitrary precision with constant cost per token by decomposing Taylor expansion into symmetric chains of tensor products, achieving orders-of-magnitude efficiency gains.

arXiv:2602.00294v1 Announce Type: cross Abstract: The most widely used artificial intelligence (AI) models today are Transformers employing self-attention. In its standard form, self-attention incurs costs that increase with context length, driving demand for storage, compute, and energy that is now outstripping society's ability to provide them. To help address this issue, we show that self-attention is efficiently computable to arbitrary precision with constant cost per token, achieving order
EfficiencyTransformersSelf-AttentionMathematical Foundations
Research arXiv (Artificial Intelligence) Feb 3

Training LLMs with Fault Tolerant HSDP on 100,000 GPUs

By Omkar Salpekar, Rohan Varma, Kenny Yu, Vladimir Ivanov, Yang Wang, Ahmed Sharif, Min Si, Shawn Xu, Feng Tian, Shengbao Zheng, Tristan Rice, Ankush Garg, Shangfu Peng, Shreyas Siravara, Wenyin Fu, Rodrigo de Castro, Adithya Gangidi, Andrey Obraztsov, Sharan Narang, Sergey Edunov, Maxim Naumov, Chunqiang Tang, Mathew Oldham

90 score
AI Analysis

Introduces Fault Tolerant HSDP for training on 100K+ GPUs, allowing individual data-parallel replicas to restart on failure while others continue. Includes novel fault-tolerant all-reduce protocol.

arXiv:2602.00277v1 Announce Type: cross Abstract: Large-scale training systems typically use synchronous training, requiring all GPUs to be healthy simultaneously. In our experience training on O(100K) GPUs, synchronous training results in a low efficiency due to frequent failures and long recovery time. To address this problem, we propose a novel training paradigm, Fault Tolerant Hybrid-Shared Data Parallelism (FT-HSDP). FT-HSDP uses data parallel replicas as units of fault tolerance. When f
Large-Scale TrainingSystemsFault ToleranceDistributed Computing
Research arXiv (Computation and Language) Feb 3

Kimi K2.5: Visual Agentic Intelligence

By Kimi Team, Tongtong Bai, Yifan Bai, Yiping Bao, S. H. Cai, Yuan Cao, Y. Charles, H. S. Che, Cheng Chen, Guanduo Chen, Huarong Chen, Jia Chen, Jiahao Chen, Jianlong Chen, Jun Chen, Kefan Chen, Liang Chen, Ruijue Chen, Xinhao Chen, Yanru Chen, Yanxu Chen, Yicun Chen, Yimin Chen, Yingjiang Chen, Yuankun Chen, Yujie Chen, Yutian Chen, Zhirong Chen, Ziwei Chen, Dazhi Cheng, Minghan Chu, Jialei Cui, Jiaqi Deng, Muxi Diao, Hao Ding, Mengfan Dong, Mengnan Dong, Yuxin Dong, Yuhao Dong, Angang Du, Chenzhuang Du, Dikang Du, Lingxiao Du, Yulun Du, Yu Fan, Shengjun Fang, Qiulin Feng, Yichen Feng, Garimugai Fu, Kelin Fu, Hongcheng Gao, Tong Gao, Yuyao Ge, Shangyi Geng, Chengyang Gong, Xiaochen Gong, Zhuoma Gongque, Qizheng Gu, Xinran Gu, Yicheng Gu, Longyu Guan, Yuanying Guo, Xiaoru Hao, Weiran He, Wenyang He, Yunjia He, Chao Hong, Hao Hu, Jiaxi Hu, Yangyang Hu, Zhenxing Hu, Ke Huang, Ruiyuan Huang, Weixiao Huang, Zhiqi Huang, Tao Jiang, Zhejun Jiang, Xinyi Jin, Yu Jing, Guokun Lai, Aidi Li, C. Li, Cheng Li, Fang Li, Guanghe Li, Guanyu Li, Haitao Li, Haoyang Li, Jia Li, Jingwei Li, Junxiong Li, Lincan Li, Mo Li, Weihong Li, Wentao Li, Xinhang Li, Xinhao Li, Yang Li, Yanhao Li, Yiwei Li, Yuxiao Li, Zhaowei Li, Zheming Li, Weilong Liao, Jiawei Lin, Xiaohan Lin, Zhishan Lin, Zichao Lin, Cheng Liu, Chenyu Liu, Hongzhang Liu, Liang Liu, Shaowei Liu, Shudong Liu, Shuran Liu, Tianwei Liu, Tianyu Liu, Weizhou Liu, Xiangyan Liu, Yangyang Liu, Yanming Liu, Yibo Liu, Yuanxin Liu, Yue Liu, Zhengying Liu, Zhongnuo Liu, Enzhe Lu, Haoyu Lu, Zhiyuan Lu, Junyu Luo, Tongxu Luo, Yashuo Luo, Long Ma, Yingwei Ma, Shaoguang Mao, Yuan Mei, Xin Men, Fanqing Meng, Zhiyong Meng, Yibo Miao, Minqing Ni, Kun Ouyang, Siyuan Pan, Bo Pang, Yuchao Qian, Ruoyu Qin, Zeyu Qin, Jiezhong Qiu, Bowen Qu, Zeyu Shang, Youbo Shao, Tianxiao Shen, Zhennan Shen, Juanfeng Shi, Lidong Shi, Shengyuan Shi, Feifan Song, Pengwei Song, Tianhui Song, Xiaoxi Song, Hongjin Su, Jianlin Su, Zhaochen Su, Lin Sui, Jinsong Sun, Junyao Sun, Tongyu Sun, Flood Sung, Yunpeng Tai, Chuning Tang, Heyi Tang, Xiaojuan Tang, Zhengyang Tang, Jiawen Tao, Shiyuan Teng, Chaoran Tian, Pengfei Tian, Ao Wang, Bowen Wang, Chensi Wang, Chuang Wang, Congcong Wang, Dingkun Wang, Dinglu Wang, Dongliang Wang, Feng Wang, Hailong Wang, Haiming Wang, Hengzhi Wang, Huaqing Wang, Hui Wang, Jiahao Wang, Jinhong Wang, Jiuzheng Wang, Kaixin Wang, Linian Wang, Qibin Wang, Shengjie Wang, Shuyi Wang, Si Wang, Wei Wang, Xiaochen Wang, Xinyuan Wang, Yao Wang, Yejie Wang, Yipu Wang, Yiqin Wang, Yucheng Wang, Yuzhi Wang, Zhaoji Wang, Zhaowei Wang, Zhengtao Wang, Zhexu Wang, Zihan Wang, Zizhe Wang, Chu Wei, Ming Wei, Chuan Wen, Zichen Wen, Chengjie Wu, Haoning Wu, Junyan Wu, Rucong Wu, Wenhao Wu, Yuefeng Wu, Yuhao Wu, Yuxin Wu, Zijian Wu, Chenjun Xiao, Jin Xie, Xiaotong Xie, Yuchong Xie, Yifei Xin, Bowei Xing, Boyu Xu, Jianfan Xu, Jing Xu, Jinjing Xu, L. H. Xu, Lin Xu, Suting Xu, Weixin Xu, Xinbo Xu, Xinran Xu, Yangchuan Xu, Yichang Xu, Yuemeng Xu, Zelai Xu, Ziyao Xu, Junjie Yan, Yuzi Yan, Guangyao Yang, Hao Yang, Junwei Yang, Kai Yang, Ningyuan Yang, Ruihan Yang, Xiaofei Yang, Xinlong Yang, Ying Yang, Yi Yang, Yi Yang, Zhen Yang, Zhilin Yang, Zonghan Yang, Haotian Yao, Dan Ye, Wenjie Ye, Zhuorui Ye, Bohong Yin, Chengzhen Yu, Longhui Yu, Tao Yu, Tianxiang Yu, Enming Yuan, Mengjie Yuan, Xiaokun Yuan, Yang Yue, Weihao Zeng, Dunyuan Zha, Haobing Zhan, Dehao Zhang, Hao Zhang, Jin Zhang, Puqi Zhang, Qiao Zhang, Rui Zhang, Xiaobin Zhang, Y. Zhang, Yadong Zhang, Yangkun Zhang, Yichi Zhang, Yizhi Zhang, Yongting Zhang, Yu Zhang, Yushun Zhang, Yutao Zhang, Yutong Zhang, Zheng Zhang, Chenguang Zhao, Feifan Zhao, Jinxiang Zhao, Shuai Zhao, Xiangyu Zhao, Yikai Zhao, Zijia Zhao, Huabin Zheng, Ruihan Zheng, Shaojie Zheng, Tengyang Zheng, Junfeng Zhong, Longguang Zhong, Weiming Zhong, M. Zhou, Runjie Zhou, Xinyu Zhou, Zaida Zhou, Jinguo Zhu, Liya Zhu, Xinhao Zhu, Yuxuan Zhu, Zhen Zhu, Jingze Zhuang, Weiyu Zhuang, Ying Zou, Xinxing Zu

88 score
AI Analysis

Kimi K2.5 is an open-source multimodal agentic model featuring joint text-vision optimization and Agent Swarm—a parallel agent orchestration framework that dynamically decomposes complex tasks. Claims SOTA across coding, vision, reasoning, and agentic tasks.

arXiv:2602.02276v1 Announce Type: new Abstract: We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel ag
Multimodal ModelsAgentsOpen SourceSOTA
Research arXiv (Artificial Intelligence) Feb 3

ReasoningBomb: A Stealthy Denial-of-Service Attack by Inducing Pathologically Long Reasoning in Large Reasoning Models

By Xiaogeng Liu, Xinyan Wang, Yechao Zhang, Sanjay Kariyappa, Chong Xiang, Muhao Chen, G. Edward Suh, Chaowei Xiao

85 score
AI Analysis

Introduces ReasoningBomb, a new class of denial-of-service attacks targeting large reasoning models by inducing pathologically long reasoning traces. Formalizes PI-DoS attacks with three key properties: amplification, stealthiness, and optimizability.

arXiv:2602.00154v1 Announce Type: cross Abstract: Large reasoning models (LRMs) extend large language models with explicit multi-step reasoning traces, but this capability introduces a new class of prompt-induced inference-time denial-of-service (PI-DoS) attacks that exploit the high computational cost of reasoning. We first formalize inference cost for LRMs and define PI-DoS, then prove that any practical PI-DoS attack should satisfy three properties: (1) a high amplification ratio, where each
AI SafetySecurityReasoning ModelsAdversarial Attacks
Research arXiv (Machine Learning) Feb 3

No Global Plan in Chain-of-Thought: Uncover the Latent Planning Horizon of LLMs

By Liyan Xu, Mo Yu, Fandong Meng, Jie Zhou

85 score
AI Analysis

Proposes Tele-Lens probing method revealing LLMs exhibit myopic planning horizon in Chain-of-Thought, conducting incremental transitions without precise global planning.

arXiv:2602.02103v1 Announce Type: new Abstract: This work stems from prior complementary observations on the dynamics of Chain-of-Thought (CoT): Large Language Models (LLMs) is shown latent planning of subsequent reasoning prior to CoT emergence, thereby diminishing the significance of explicit CoT; whereas CoT remains critical for tasks requiring multi-step reasoning. To deepen the understanding between LLM's internal states and its verbalized reasoning trajectories, we investigate the latent
LLM ReasoningChain-of-ThoughtInterpretabilityPlanning
Research arXiv (Machine Learning) Feb 3

AICD Bench: A Challenging Benchmark for AI-Generated Code Detection

By Daniil Orel, Dilshod Azizov, Indraneil Paul, Yuxia Wang, Iryna Gurevych, Preslav Nakov

84 score
AI Analysis

Introduces AICD Bench, comprehensive benchmark for AI-generated code detection with 2M examples, 77 models across 11 families, 9 languages, including reasoning models and three realistic detection tasks.

arXiv:2602.02079v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly capable of generating functional source code, raising concerns about authorship, accountability, and security. While detecting AI-generated code is critical, existing datasets and benchmarks are narrow, typically limited to binary human-machine classification under in-distribution settings. To bridge this gap, we introduce $\emph{AICD Bench}$, the most comprehensive benchmark for AI-generated code dete
AI-Generated Content DetectionCode GenerationBenchmarkAI Safety
Research arXiv (Artificial Intelligence) Feb 3

BLOCK-EM: Preventing Emergent Misalignment by Blocking Causal Features

By Muhammed Ustaomeroglu, Guannan Qu

83 score
AI Analysis

Proposes BLOCK-EM for preventing emergent misalignment by identifying and constraining internal features that control misaligned behavior during fine-tuning. Achieves up to 95% reduction in emergent misalignment across six domains.

arXiv:2602.00767v1 Announce Type: cross Abstract: Emergent misalignment can arise when a language model is fine-tuned on a narrowly scoped supervised objective: the model learns the target behavior, yet also develops undesirable out-of-domain behaviors. We investigate a mechanistic approach to preventing emergent misalignment by identifying a small set of internal features that reliably control the misaligned behavior and then discouraging the model from strengthening these features during fine
AI SafetyEmergent MisalignmentMechanistic InterpretabilityFine-tuning
Research arXiv (Machine Learning) Feb 3

Grad2Reward: From Sparse Judgment to Dense Rewards for Improving Open-Ended LLM Reasoning

By Zheng Zhang, Ao Lu, Yuanhao Zeng, Ziwei Shan, Jinjin Guo, Lufei Li, Yexin Li, Kan Ren

83 score
AI Analysis

Introduces Grad2Reward framework that extracts dense process rewards directly from LLM judge gradients, converting sparse sequence-level rewards into fine-grained supervision for RLHF on open-ended tasks.

arXiv:2602.01791v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has catalyzed significant breakthroughs in complex LLM reasoning within verifiable domains, such as mathematics and programming. Recent efforts have sought to extend this paradigm to open-ended tasks by employing LLMs-as-a-Judge to provide sequence-level rewards for policy optimization. However, these rewards are inherently sparse, failing to provide the fine-grained supervision necessary for g
RLHFReward ModelingLLM Reasoning
Research arXiv (Artificial Intelligence) Feb 3

MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety

By Xiaoyu Wen, Zhida He, Han Qi, Ziyu Wan, Zhongtian Ma, Ying Wen, Tianhang Zheng, Xingcheng Xu, Chaochao Lu, Qiaosheng Zhang

82 score
AI Analysis

MAGIC frames LLM safety alignment as adversarial game where attacker agent discovers vulnerabilities while defender agent learns to refuse, enabling co-evolution that uncovers long-tail vulnerabilities.

arXiv:2602.01539v1 Announce Type: new Abstract: Ensuring robust safety alignment is crucial for Large Language Models (LLMs), yet existing defenses often lag behind evolving adversarial attacks due to their \textbf{reliance on static, pre-collected data distributions}. In this paper, we introduce \textbf{MAGIC}, a novel multi-turn multi-agent reinforcement learning framework that formulates LLM safety alignment as an adversarial asymmetric game. Specifically, an attacker agent learns to iterati
AI SafetyAdversarial LearningReinforcement LearningAlignment
Research arXiv (Artificial Intelligence) Feb 3

Breaking the Reversal Curse in Autoregressive Language Models via Identity Bridge

By Xutao Ma, Yixiao Huang, Hanlin Zhu, Somayeh Sojoudi

82 score
AI Analysis

Proposes 'Identity Bridge' to address the reversal curse in autoregressive LLMs, where models trained on 'A→B' cannot deduce 'B→A'. Claims to mitigate what was previously considered a fundamental limitation of causal LLMs.

arXiv:2602.02470v1 Announce Type: new Abstract: Autoregressive large language models (LLMs) have achieved remarkable success in many complex tasks, yet they can still fail in very simple logical reasoning such as the "reversal curse" -- when trained on forward knowledge data of the form "$A \rightarrow B$" (e.g., Alice's husband is Bob), the model is unable to deduce the reversal knowledge "$B \leftarrow A$" (e.g., Bob's wife is Alice) during test. Extensive prior research suggests that this fa
Language ModelsReasoningLLM Limitations
Research arXiv (Artificial Intelligence) Feb 3

Learning Robust Reasoning through Guided Adversarial Self-Play

By Shuozhe Li, Vaishnav Tadiparthi, Kwonjoon Lee, Nakul Agarwal, Hossein Nourkhiz Mahjoub, Ehsan Moradi Pari, Lizhang Chen, Amy Zhang, Liu Leqi

82 score
AI Analysis

GASP introduces adversarial self-play within a single model to train detect-and-repair capabilities for reasoning, using only outcome verification. A polluter induces coherent corruptions while an agent learns to recover.

arXiv:2602.00173v1 Announce Type: cross Abstract: Reinforcement learning from verifiable rewards (RLVR) produces strong reasoning models, yet they can fail catastrophically when the conditioning context is fallible (e.g., corrupted chain-of-thought, misleading partial solutions, or mild input perturbations), since standard RLVR optimizes final-answer correctness only under clean conditioning. We introduce GASP (Guided Adversarial Self-Play), a robustification method that explicitly trains detec
ReasoningReinforcement LearningRobustnessSelf-Play
Research arXiv (Artificial Intelligence) Feb 3

Jailbreaking LLMs via Calibration

By Yuxuan Lu, Yongkang Guo, Yuqing Kong

82 score
AI Analysis

Models safety alignment in LLMs as a systematic distortion of pre-alignment distributions and casts jailbreaking as a forecast aggregation problem. Derives optimal aggregation strategy and shows logit-arithmetic methods are a special case, proposing broader family of attacks.

arXiv:2602.00619v1 Announce Type: cross Abstract: Safety alignment in Large Language Models (LLMs) often creates a systematic discrepancy between a model's aligned output and the underlying pre-aligned data distribution. We propose a framework in which the effect of safety alignment on next-token prediction is modeled as a systematic distortion of a pre-alignment distribution. We cast Weak-to-Strong Jailbreaking as a forecast aggregation problem and derive an optimal aggregation strategy charac
AI SafetyJailbreakingLLM AlignmentSecurity