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
63 results for “keyword:"refusal"” in ICLR 2027. 17 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]KV Compression Cannot Double as a Jailbreak Defense: The Harmfulness-then-Refusal Trajectory Guides Token Deletion▼5.8k2.33
[PDF]A Harmful Request Is Still a Request: Benign Fine-Tuning Rewrites How Aligned Models Begin to Answer▲4012.67
[PDF]A Jailbreak Success Coordinate Outside the Refusal Subspace: Decodable Before Generation, Reachable from the Input▼5.8k2.33
[PDF]Safety Fragility Is a Routing Problem: The Attention Bottleneck Between Harm Detection and Refusal▲1.9k3.67
[PDF]A Linear Name-Origin Prior in the Recall Heads: How Language Models Fabricate and Misrecall Facts About People▲5103.00
[PDF]Anchored but Not Protected: When Refusal Erodes During Fine-Tuning, and Whether Holding One Direction Stops It▲13k2.33
[PDF]Attainment Is Not Stationarity: Objective and Optimizer Sources of Behavioral Drift after Safety Alignment▼4.3k2.00
[PDF]When Conversation is the Attack: Mining Transferable Contexts for Harmful Compliance and Over-Refusal▲3.0k2.33
[PDF]StiefelSteer: Invertible Refusal Steering via Low-Dimensional Rotations of the Residual Stream▼1.2k1.67
[PDF]GroundedMem: Source-Resolved Retrieval and Conservative Refusal for Long-Term Conversational Memory▲3.0k2.33
[PDF]Beyond Refuse-or-Comply: Fine-Grained Policy-to-Action Grounding in Safety-Aligned Language Models▼5.8k2.33
[PDF]Saturated Screening, Concentrated Cost: What a Full-Stack Ablation Table Reads in a Redundant Refusal Stack▼1.2k1.67
[PDF]Frozen Input-Harm Probes under Compression: Retained AUROC Does Not Mean Retained Refusal▼1.2k1.67
[PDF]Before the Refusal: Pre-Response Activation Signatures of False Refusal in Safety-Aligned Language Models▼1.2k1.67
[PDF]TotalGuardEval: Benchmark for Evaluating Guardrail Models on Underrepresented Risk Categories▼5.8k2.33
[PDF]RefusalTrace: Localizing and Validating the Internal Refusal Pathway of LLMs with Sparse Autoencoder Features▲9.7k2.75
Withdrawn or desk rejected (17)
[PDF]Why Benign Post-Training Induces Safety Gate Collapse: Changes in Harmful Activation Geometry Relative to the Inherited Refusal GateWithdrawn1.67
[PDF]The Simulator Says Yes: Language World Models Narrate the Executions Their Environment RefusedDesk rejected
[PDF]Faithfulness Anchoring as Defense Against CoT Hijacking: A Mechanism-First Approach via Residual-Stream AnalysisWithdrawn1.33
[PDF]Easier Said Than Done: Strangers, Impostors, and the Negatives That Teach a Small Learner to Say "I Don't Know"Desk rejected