{"total":68,"rows":[{"id":"2506.10805","title":"Detecting High-Stakes Interactions with Activation Probes","created":"2025-06-12","cited_by_count":53,"area":null},{"id":"2509.10625","title":"No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes","created":"2025-09-12","cited_by_count":38,"area":"mechinterp"},{"id":"2507.01786","title":"Probing and Steering Evaluation Awareness of Language Models","created":"2025-07-02","cited_by_count":35,"area":null},{"id":"2504.04072","title":"Among Us: A Sandbox for Measuring and Detecting Agentic Deception","created":"2025-04-05","cited_by_count":26,"area":"agents"},{"id":"2510.04040","title":"FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning","created":"2025-10-05","cited_by_count":26,"area":"lr"},{"id":"2506.14261","title":"RL-Obfuscation: Can Language Models Learn to Evade Latent-Space Monitors?","created":"2025-06-17","cited_by_count":24,"area":null},{"id":"2512.08093","title":"Training LLMs for Honesty via Confessions","created":"2025-12-08","cited_by_count":22,"area":null},{"id":"2502.17601","title":"Representation Engineering for Large-Language Models: Survey and Research Challenges","created":"2025-02-24","cited_by_count":21,"area":"mechinterp"},{"id":"2605.11887","title":"Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models","created":"2026-05-12","cited_by_count":21,"area":"mechinterp"},{"id":"2507.12691","title":"Benchmarking Deception Probes via Black-to-White Performance Boosts","created":"2025-07-16","cited_by_count":19,"area":null},{"id":"2509.26238","title":"Beyond Linear Probes: Dynamic Safety Monitoring for Language Models","created":"2025-09-30","cited_by_count":16,"area":"safety"},{"id":"2504.20271","title":"Investigating task-specific prompts and sparse autoencoders for activation monitoring","created":"2025-04-28","cited_by_count":15,"area":"mechinterp"},{"id":"2511.06626","title":"Spilling the Beans: Teaching LLMs to Self-Report Their Hidden Objectives","created":"2025-11-10","cited_by_count":15,"area":"agents"},{"id":"2504.03255","title":"Inherent and emergent liability issues in LLM-based agentic systems: a principal-agent perspective","created":"2025-04-04","cited_by_count":13,"area":"agents"},{"id":"2511.22662","title":"Difficulties with Evaluating a Deception Detector for AIs","created":"2025-11-27","cited_by_count":13,"area":null},{"id":"2602.15515","title":"The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes","created":"2026-02-17","cited_by_count":12,"area":null},{"id":"2507.15886","title":"Combining Cost-Constrained Runtime Monitors for AI Safety","created":"2025-07-19","cited_by_count":10,"area":null},{"id":"2505.13787","title":"Preference Learning with Lie Detectors can Induce Honesty or Evasion","created":"2025-05-20","cited_by_count":9,"area":"alignment"},{"id":"2509.00591","title":"Probe-Rewrite-Evaluate: A Workflow for Reliable Benchmarks and Quantifying Evaluation Awareness","created":"2025-08-30","cited_by_count":9,"area":null},{"id":"2511.17408","title":"The Impact of Off-Policy Training Data on Probe Generalisation","created":"2025-11-21","cited_by_count":9,"area":null},{"id":"2512.11949","title":"Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitors","created":"2025-12-12","cited_by_count":9,"area":"lr"},{"id":"2502.14445","title":"PredictaBoard: Benchmarking LLM Score Predictability","created":"2025-02-20","cited_by_count":7,"area":null},{"id":"2509.03518","title":"Can LLMs Lie? Investigation beyond Hallucination","created":"2025-09-03","cited_by_count":7,"area":"mechinterp"},{"id":"2604.01151","title":"Detecting Multi-Agent Collusion Through Multi-Agent Interpretability","created":"2026-04-01","cited_by_count":7,"area":"agents"},{"id":"2508.19505","title":"Caught in the Act: a mechanistic approach to detecting deception","created":"2025-08-27","cited_by_count":6,"area":null},{"id":"2510.05192","title":"From surveillance to signalling: escalation channels as environmental controls for agentic AI","created":"2025-10-06","cited_by_count":6,"area":"agents"},{"id":"2602.22303","title":"Training Agents to Self-Report Misbehavior","created":"2026-02-25","cited_by_count":6,"area":"agents"},{"id":"2603.17199","title":"Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probing","created":"2026-03-17","cited_by_count":6,"area":"lr"},{"id":"2605.15377","title":"Ensemble Monitoring for AI Control: Diverse Signals Outweigh More Compute","created":"2026-05-14","cited_by_count":6,"area":"agents"},{"id":"2606.13870","title":"Mirage Probes: How Vision Models Fake Visual Understanding","created":"2026-06-11","cited_by_count":5,"area":"mllm"},{"id":"2508.05625","title":"How Do LLMs Persuade? Linear Probes Can Uncover Persuasion Dynamics in Multi-Turn Conversations","created":"2025-08-07","cited_by_count":4,"area":"mechinterp"},{"id":"2604.13386","title":"Linear Probe Accuracy Scales with Model Size and Benefits from Multi-Layer Ensembling","created":"2026-04-15","cited_by_count":4,"area":"alignment"},{"id":"2606.12618","title":"\"Did you lie?\" Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms","created":"2026-06-10","cited_by_count":4,"area":"lr"},{"id":"2504.07831","title":"Deceptive Automated Interpretability: Language Models Coordinating to Fool Oversight Systems","created":"2025-04-10","cited_by_count":3,"area":"mechinterp"},{"id":"2512.24560","title":"Localized Calibrated Uncertainty in Code Language Models","created":"2025-12-31","cited_by_count":3,"area":null},{"id":"2603.26635","title":"Deception and Communication in Autonomous Multi-Agent Systems: An Experimental Study with Among Us","created":"2026-03-27","cited_by_count":3,"area":"agents"},{"id":"2604.03121","title":"An Independent Safety Evaluation of Kimi K2.5","created":"2026-04-03","cited_by_count":3,"area":"agents"},{"id":"2605.13339","title":"Probing Persona-Dependent Preferences in Language Models","created":"2026-05-13","cited_by_count":3,"area":null},{"id":"2509.21344","title":"Linear probes rely on textual evidence: Results from leakage mitigation studies in language models","created":"2025-09-16","cited_by_count":2,"area":"mechinterp"},{"id":"2509.22739","title":"Painless Activation Steering: An Automated, Lightweight Approach for Post-Training Large Language Models","created":"2025-09-25","cited_by_count":2,"area":"mechinterp"},{"id":"2510.21531","title":"Probe-based Fine-tuning for Reducing Toxicity","created":"2025-10-24","cited_by_count":2,"area":null},{"id":"2604.00209","title":"Do LLMs Know What Is Private Internally? Probing and Steering Contextual Privacy Norms in Large Language Model Representations","created":"2026-03-31","cited_by_count":2,"area":null},{"id":"2605.15164","title":"Position: Behavioural Assurance Cannot Verify the Safety Claims Governance Now Demands","created":"2026-05-14","cited_by_count":2,"area":null},{"id":"2605.27958","title":"Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations","created":"2026-05-27","cited_by_count":2,"area":"mechinterp"},{"id":"2606.17478","title":"Decoding Hidden Deception in Reasoning LLMs: Activation Explainers for Deception Auditing","created":"2026-06-16","cited_by_count":2,"area":"lr"},{"id":"2609.24967","title":"Emergent Collusion in Long-Horizon LLM Agent Interaction","created":"2026-09-21","cited_by_count":2,"area":"agents"},{"id":"2510.06527","title":"Wide Neural Networks as a Baseline for the Computational No-Coincidence Conjecture","created":"2025-10-08","cited_by_count":1,"area":null},{"id":"2511.00554","title":"Red-teaming Activation Probes using Prompted LLMs","created":"2025-11-01","cited_by_count":1,"area":"safety"},{"id":"2603.25861","title":"Why Safety Probes Catch Liars But Miss Fanatics","created":"2026-03-26","cited_by_count":1,"area":null},{"id":"2605.16339","title":"Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders","created":"2026-05-07","cited_by_count":1,"area":"alignment"},{"id":"2605.25310","title":"Tool-Call Dependency Structure is Linearly Decodable in LLM Agent Residual Streams","created":"2026-05-25","cited_by_count":1,"area":"agents"},{"id":"2606.24251","title":"Probing the Misaligned Thinking Process of Language Models","created":"2026-06-23","cited_by_count":1,"area":"lr"},{"id":"2607.01047","title":"Conversable Complexity: Agentic LLM Collectives as Interpretable Substrates","created":"2026-07-01","cited_by_count":1,"area":"agents"},{"id":"2607.13346","title":"The Refusal Residue: When Probes Catch Alignment Faking and When They Don't","created":"2026-07-15","cited_by_count":1,"area":null},{"id":"2506.04774","title":"Fine-Grained Interpretation of Political Opinions in Large Language Models","created":"2025-06-05","cited_by_count":0,"area":"mechinterp"},{"id":"2507.10106","title":"BlueGlass: A Framework for Composite AI Safety","created":"2025-07-14","cited_by_count":0,"area":null},{"id":"2603.10003","title":"Probing the Limits of the Lie Detector Approach to LLM Deception","created":"2026-02-16","cited_by_count":0,"area":null},{"id":"2604.24801","title":"Architecture Determines Observability of Transformers","created":"2026-04-27","cited_by_count":0,"area":null},{"id":"2605.09391","title":"Do Linear Probes Generalize Better in Persona Coordinates?","created":"2026-05-10","cited_by_count":0,"area":"mechinterp"},{"id":"2607.14791","title":"Transcoders for Investigating Deception in Language Models","created":"2026-07-16","cited_by_count":0,"area":"mechinterp"},{"id":"2608.00583","title":"A False Average: Pooled CoT-Monitor Accuracy Conceals a Reasoning-Dependent Fragility","created":"2026-08-01","cited_by_count":0,"area":"lr"},{"id":"2609.00180","title":"Asymmetries in Spontaneous and Instructed Deception","created":"2026-08-31","cited_by_count":0,"area":null},{"id":"2609.02893","title":"Probe Generalization as Subspace Selection for OOD Deception Detection","created":"2026-07-01","cited_by_count":0,"area":"mechinterp"},{"id":"2609.04721","title":"Locating and Steering Refusal Beyond Attention","created":"2026-09-04","cited_by_count":0,"area":null},{"id":"2609.10739","title":"The Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth Probes","created":"2026-09-09","cited_by_count":0,"area":"mechinterp"},{"id":"2609.31796","title":"Same Probe, Different Numbers: Are Activation Probes Robust to Inference-Time Numerical Non-Determinism?","created":"2026-09-25","cited_by_count":0,"area":null},{"id":"2610.10594","title":"Coverage, Not Difficulty, Sets How Much Synthetic Data an Activation Probe Needs","created":"2026-10-06","cited_by_count":0,"area":"synth"},{"id":"2610.20179","title":"Cross-Model In-Context Neurofeedback Reveals (Limited) Activation Control in LLMs","created":"2026-10-14","cited_by_count":0,"area":null}],"next_cursor":null}