惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

S
Schneier on Security
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
T
Threat Research - Cisco Blogs
C
Cyber Attacks, Cyber Crime and Cyber Security
C
CXSECURITY Database RSS Feed - CXSecurity.com
A
Arctic Wolf
Security Latest
Security Latest
Simon Willison's Weblog
Simon Willison's Weblog
I
Intezer
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
T
Troy Hunt's Blog
Latest news
Latest news
Help Net Security
Help Net Security
S
Security Affairs
Webroot Blog
Webroot Blog
The Hacker News
The Hacker News
AI
AI
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
T
Tor Project blog
Forbes - Security
Forbes - Security
Google DeepMind News
Google DeepMind News
AWS News Blog
AWS News Blog
Attack and Defense Labs
Attack and Defense Labs
P
Proofpoint News Feed
www.infosecurity-magazine.com
www.infosecurity-magazine.com
H
Help Net Security
L
Lohrmann on Cybersecurity
S
SegmentFault 最新的问题
Google Online Security Blog
Google Online Security Blog
MongoDB | Blog
MongoDB | Blog
Cyberwarzone
Cyberwarzone
The Last Watchdog
The Last Watchdog
S
Securelist
N
News and Events Feed by Topic
S
Secure Thoughts
F
Fortinet All Blogs
博客园_首页
C
Cybersecurity and Infrastructure Security Agency CISA
量子位
M
MIT News - Artificial intelligence
F
Full Disclosure
T
The Blog of Author Tim Ferriss
T
Tailwind CSS Blog
Recent Commits to openclaw:main
Recent Commits to openclaw:main
Microsoft Security Blog
Microsoft Security Blog
I
InfoQ
P
Privacy International News Feed
L
LangChain Blog
Know Your Adversary
Know Your Adversary
C
CERT Recently Published Vulnerability Notes

cs.CL updates on arXiv.org

Playing Along: Learning a Double-Agent Defender for Belief Steering via Theory of Mind RPA-Check: A Multi-Stage Automated Framework for Evaluating Dynamic LLM-based Role-Playing Agents A Triadic Suffix Tokenization Scheme for Numerical Reasoning Hidden Measurement Error in LLM Pipelines Distorts Annotation, Evaluation, and Benchmarking Synthius-Mem: Brain-Inspired Hallucination-Resistant Persona Memory Achieving 94.4% Memory Accuracy and 99.6% Adversarial Robustness on LoCoMo Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory NovBench: Evaluating Large Language Models on Academic Paper Novelty Assessment Policy Split: Incentivizing Dual-Mode Exploration in LLM Reinforcement with Dual-Mode Entropy Regularization METER: Evaluating Multi-Level Contextual Causal Reasoning in Large Language Models Think Before you Write: QA-Guided Reasoning for Character Descriptions in Books METRO: Towards Strategy Induction from Expert Dialogue Transcripts for Non-collaborative Dialogues Retrieval as Generation: A Unified Framework with Self-Triggered Information Planning Do LLMs Know Tool Irrelevance? Demystifying Structural Alignment Bias in Tool Invocations Enhancing Multimodal Large Language Models for Ancient Chinese Character Evolution Analysis via Glyph-Driven Fine-Tuning Polyglot Teachers: Evaluating Language Models for Multilingual Synthetic Data Generation Exploring Knowledge Conflicts for Faithful LLM Reasoning: Benchmark and Method CocoaBench: Evaluating Unified Digital Agents in the Wild MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis Evaluating Memory Capability in Continuous Lifelog Scenario How Robust Are Large Language Models for Clinical Numeracy? An Empirical Study on Numerical Reasoning Abilities in Clinical Contexts Efficient Training for Cross-lingual Speech Language Models Shared Emotion Geometry Across Small Language Models: A Cross-Architecture Study of Representation, Behavior, and Methodological Confounds A Systematic Analysis of the Impact of Persona Steering on LLM Capabilities Uncertainty-Aware Web-Conditioned Scientific Fact-Checking When Valid Signals Fail: Regime Boundaries Between LLM Features and RL Trading Policies When Verification Fails: How Compositionally Infeasible Claims Escape Rejection Mem$^2$Evolve: Towards Self-Evolving Agents via Co-Evolutionary Capability Expansion and Experience Distillation AOP-Smart: A RAG-Enhanced Large Language Model Framework for Adverse Outcome Pathway Analysis OccuBench: Evaluating AI Agents on Real-World Professional Tasks via Language Environment Simulation Advancing Polish Language Modeling through Tokenizer Optimization in the Bielik v3 7B and 11B Series TInR: Exploring Tool-Internalized Reasoning in Large Language Models Do BERT Embeddings Encode Narrative Dimensions? A Token-Level Probing Analysis of Time, Space, Causality, and Character in Fiction Generating Multiple-Choice Knowledge Questions with Interpretable Difficulty Estimation using Knowledge Graphs and Large Language Models Deep-Reporter: Deep Research for Grounded Multimodal Long-Form Generation BlasBench: An Open Benchmark for Irish Speech Recognition Too Nice to Tell the Truth: Quantifying Agreeableness-Driven Sycophancy in Role-Playing Language Models Learning and Enforcing Context-Sensitive Control for LLMs HeceTokenizer: A Syllable-Based Tokenization Approach for Turkish Retrieval Efficient Process Reward Modeling via Contrastive Mutual Information Computational Lesions in Multilingual Language Models Separate Shared and Language-specific Brain Alignment Bridging Linguistic Gaps: Cross-Lingual Mapping in Pre-Training and Dataset for Enhanced Multilingual LLM Performance Knowing What to Stress: A Discourse-Conditioned Text-to-Speech Benchmark Early Decisions Matter: Proximity Bias and Initial Trajectory Shaping in Non-Autoregressive Diffusion Language Models LLMs Should Incorporate Explicit Mechanisms for Human Empathy ReFEree: Reference-Free and Fine-Grained Method for Evaluating Factual Consistency in Real-World Code Summarization Structure-Grounded Knowledge Retrieval via Code Dependencies for Multi-Step Data Reasoning From Query to Counsel: Structured Reasoning with a Multi-Agent Framework and Dataset for Legal Consultation Dynamic Adaptive Attention and Supervised Contrastive Learning: A Novel Hybrid Framework for Text Sentiment Classification EviCare: Enhancing Diagnosis Prediction with Deep Model-Guided Evidence for In-Context Reasoning NOSE: Neural Olfactory-Semantic Embedding with Tri-Modal Orthogonal Contrastive Learning Instruction Data Selection via Answer Divergence CodaRAG: Connecting the Dots with Associativity Inspired by Complementary Learning Turing or Cantor: That is the Question LASQ: A Low-resource Aspect-based Sentiment Quadruple Extraction Dataset NameBERT: Scaling Name-Based Nationality Classification with LLM-Augmented Open Academic Data BLUEmed: Retrieval-Augmented Multi-Agent Debate for Clinical Error Detection A Structured Clustering Approach for Inducing Media Narratives Adaptive Multi-Expert Reasoning via Difficulty-Aware Routing and Uncertainty-Guided Aggregation Comparative Analysis of Large Language Models in Healthcare CodeComp: Structural KV Cache Compression for Agentic Coding Relational Probing: LM-to-Graph Adaptation for Financial Prediction FAITH: Factuality Alignment through Integrating Trustworthiness and Honestness ODUTQA-MDC: A Task for Open-Domain Underspecified Tabular QA with Multi-turn Dialogue-based Clarification Nationality encoding in language model hidden states: Probing culturally differentiated representations in persona-conditioned academic text Think in Sentences: Explicit Sentence Boundaries Enhance Language Model's Capabilities Training-Free Cross-Lingual Dysarthria Severity Assessment via Phonological Subspace Analysis in Self-Supervised Speech Representations CircuitSynth: Reliable Synthetic Data Generation Who Wrote This Line? Evaluating the Detection of LLM-Generated Classical Chinese Poetry SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models Why Supervised Fine-Tuning Fails to Learn: A Systematic Study of Incomplete Learning in Large Language Models Reason Only When Needed: Efficient Generative Reward Modeling via Model-Internal Uncertainty ASPIRin: Action Space Projection for Interactivity-Optimized Reinforcement Learning in Full-Duplex Speech Language Models Mirroring Minds: Asymmetric Linguistic Accommodation and Diagnostic Identity in ADHD and Autism Reddit Communities Computational Implementation of a Model of Category-Theoretic Metaphor Comprehension CoSToM:Causal-oriented Steering for Intrinsic Theory-of-Mind Alignment in Large Language Models Weird Generalization is Weirdly Brittle Human vs. Machine Deception: Distinguishing AI-Generated and Human-Written Fake News Using Ensemble Learning Should We be Pedantic About Reasoning Errors in Machine Translation? Simulating Organized Group Behavior: New Framework, Benchmark, and Analysis Spoiler Alert: Narrative Forecasting as a Metric for Tension in LLM Storytelling Claim2Vec: Embedding Fact-Check Claims for Multilingual Similarity and Clustering GIANTS: Generative Insight Anticipation from Scientific Literature Many-Tier Instruction Hierarchy in LLM Agents Interactive ASR: Towards Human-Like Interaction and Semantic Coherence Evaluation for Agentic Speech Recognition Regime-Conditional Retrieval: Theory and a Transferable Router for Two-Hop QA ASTRA: Adaptive Semantic Tree Reasoning Architecture for Complex Table Question Answering Revisiting the Capacity Gap in Chain-of-Thought Distillation from a Practical Perspective Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs $p1$: Better Prompt Optimization with Fewer Prompts Every Response Counts: Quantifying Uncertainty of LLM-based Multi-Agent Systems through Tensor Decomposition Skip-Connected Policy Optimization for Implicit Advantage PRAGMA: Revolut Foundation Model Linear Representations of Hierarchical Concepts in Language Models Generating High Quality Synthetic Data for Dutch Medical Conversations HumorGen: Cognitive Synergy for Humor Generation in Large Language Models via Persona-Based Distillation Toward Generalized Cross-Lingual Hateful Language Detection with Web-Scale Data and Ensemble LLM Annotations Self-Calibrating Language Models via Test-Time Discriminative Distillation Re-Mask and Redirect: Exploiting Denoising Irreversibility in Diffusion Language Models H-AdminSim: A Multi-Agent Simulator for Realistic Hospital Administrative Workflows with FHIR Integration Reasoning Models Will Sometimes Lie About Their Reasoning
On the Existence and Behavior of Secondary Attention Sinks
Jeffrey T. H. Wong, Cheng Zhang, Louis Mahon, Wayne Luk, Anton I · 2025-12-22 · via cs.CL updates on arXiv.org

Attention sinks are tokens, often the beginning-of-sequence (BOS) token, that receive disproportionately high attention despite limited semantic relevance. In this work, we identify a class of attention sinks, which we term secondary sinks, that differ fundamentally from the sinks studied in prior works, which we term primary sinks. While prior works have identified that tokens other than BOS can sometimes become sinks, they were found to exhibit properties analogous to the BOS token. Specifically, they emerge at the same layer, persist throughout the network and draw a large amount of attention mass. Whereas, we find the existence of secondary sinks that arise primarily in middle layers and can persist for a variable number of layers, and draw a smaller, but still significant, amount of attention mass. Through extensive experiments across 11 model families, we analyze where these secondary sinks appear, their properties, how they are formed, and their impact on the attention mechanism. Specifically, we show that: (1) these sinks are formed by specific middle-layer MLP modules; these MLPs map token representations to vectors that align with the direction of the primary sink of that layer. (2) The $\ell_2$-norm of these vectors determines the sink score of the secondary sink, and also the number of layers it lasts for, thereby leading to different impacts on the attention mechanisms accordingly. (3) The primary sink weakens in middle layers, coinciding with the emergence of secondary sinks. We observe that in larger-scale models, the location and lifetime of the sinks, together referred to as sink levels, appear in a more deterministic and frequent manner. Specifically, we identify three sink levels in QwQ-32B and six levels in Qwen3-14B. We open-sourced our findings at github.com/JeffreyWong20/Secondary-Attention-Sinks.