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
689 results for “keyword:"mechanistic interpretability"” in ICLR 2027. 142 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]"Benign-Shifted Compliance": A Mechanistic Insight into Safety Behaviors of Diffusion Language Models under Adversarial Prefixes▲5.4k3.00
[PDF]Tracing the Causal Transition from Event Relations to Moral Judgment in Language Models▼1.2k1.67
[PDF]The Evidence Anchoring Boundary: Where Visual Interventions Matter and When They Help in Multimodal LLMs▼5.8k2.33
[PDF]How Language Models Differ in Redistributing Attention-Head Activity under Serial Demand▼1.2k1.67
[PDF]Attention Tracks Reaction Centers, But Doesn't Need Them: A Causal Study of a Retrosynthesis Transformer▼4.3k2.00
[PDF]Mechanistic Interpretability Findings That Generalize: Chaperone-Dependent Folding in ESMFold2▼4.3k2.00
[PDF]From Similar Code to Neural Correspondence: Causal Interventions Guided by Program Structure▼5.8k2.33
[PDF]CTA Triplet: Disentangling Context, Constraint and Alignment Stages in In-Context Learning with Source-Selective Validation▼12k1.67
[PDF]Initial Spectral Power Predicts Frequency Selection in Grokking and Steers It in Tied MLPs▼1361.33
[PDF]A Steerable Internal Direction Links Distinct Mental-Accounting Biases in Large Language Models▼4.3k2.00
[PDF]Amplification Without Repair: Causal Limits of Activation Steering in Abstract Reasoning▼5.9k2.50
[PDF]The Behavioral Anatomy of Multi-Turn Jailbreaks: Critical Turns, Context Dependence, and Self-Conditioning▼1361.33
Withdrawn or desk rejected (142)
[PDF]Why Benign Post-Training Induces Safety Gate Collapse: Changes in Harmful Activation Geometry Relative to the Inherited Refusal GateWithdrawn1.67
[PDF]Concept Subspaces Compute Beyond the Logit Lens: A Weights-Only Test for Locating Representations Upstream of ReadoutWithdrawn