NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated
acceptodds: the papers of ICLR 2027
517 results for “keyword:"activation"” in ICLR 2027. 119 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]The Signal Survives, the Threshold Slips: Small LLM Agents After Quantization and Fine-Tuning▼1.2k1.67
[PDF]Robust Steerability: Closed-Loop Control of Language Model Representations Under Distribution Shift▼4.5k2.67
[PDF]Understanding and Improving Adversarial Answer Recovery via Activation Steering in Vision–Language Models▲3952.67
[PDF]Same Probe, Different Numbers: Are Activation Probes Robust to Inference-Time Numerical Non-Determinism?▲11k2.00
[PDF]Right Feature, Right Place: Residual-Driven Construction of Cauchy-Adaptive Physics-Informed Neural Networks▼1.2k3.00
[PDF]The Verbalization Gap: Stress-Testing Activation Monitoring with Natural Language Autoencoders▼1361.33
[PDF]Interpreting and Steering Trajectory Entrainment Behavior in Goal-Oriented Conversations▼1.2k1.67
[PDF]From Behavior to Mechanism: Tracing Divergent Response Modes in Frontier Language Models▼1361.33
[PDF]Utility-Preserving Relabeling Reveals Semantic Sensitivity in Language-Model Decision Making▼1.2k1.67
[PDF]A Linear Name-Origin Prior in the Recall Heads: How Language Models Fabricate and Misrecall Facts About People▲5103.00
[PDF]Form Before Harm: Diagnosing the Semantic Influence of Jailbreak Formats with Activation Steering▼1.2k1.67
[PDF]Right on Average, Wrong per Respondent: A Mechanistic Diagnosis of LLM-Simulated Choice▼1.2k1.67
[PDF]Context-Robust: Context-Informed Activation Calibration for Mixture-of-Experts Language Models▲14k3.00
[PDF]Evidence States Survive Where Answers Fail: Decodability and Causal Access in Language Models▼4.3k2.00
[PDF]Auditing the Auditor: Ground-Truth Controls for Natural-Language Explanations of LLM Activations▼4.3k2.00
[PDF]Anchored but Not Protected: When Refusal Erodes During Fine-Tuning, and Whether Holding One Direction Stops It▲13k2.33
[PDF]Piecewise Bernstein Networks: Scalable, Compatible and Universal Learnable Activations▼4.3k2.00
[PDF]Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized Deception▼1.1k3.25
[PDF]From Hallucinated Targets to Runaway Repetition: Termination Control in Long Structured Generation▼5.9k2.50
Withdrawn or desk rejected (119)
[PDF]Why Benign Post-Training Induces Safety Gate Collapse: Changes in Harmful Activation Geometry Relative to the Inherited Refusal GateWithdrawn1.67
[PDF]Activation Density in Transformer Feed-Forward Networks Is Set by the Nonlinearity, Not the WidthWithdrawn2.00
[PDF]Concept Subspaces Compute Beyond the Logit Lens: A Weights-Only Test for Locating Representations Upstream of ReadoutWithdrawn