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
104 results for “keyword:"safety alignment"” in ICLR 2027. 24 withdrawn or desk rejected listed apart.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▲3952.67
[PDF]Same Conversation, Different Actions: Phase-Conditioned Causal Steering for Multi-Turn LLM Safety▲1.7k3.33
[PDF]Safety Fragility Is a Routing Problem: The Attention Bottleneck Between Harm Detection and Refusal▲1.9k3.67
[PDF]Reasoning Against Alignment: How Viewpoint Processing Destabilizes Safety Boundaries in Aligned LLMs▼1.2k1.67
[PDF]Attainment Is Not Stationarity: Objective and Optimizer Sources of Behavioral Drift after Safety Alignment▼4.3k2.00
[PDF]Rethinking LLM Jailbreak Evaluation: Continuous Risk Landscapes Beyond Binary Outcomes▼1.2k1.67
[PDF]When Safety Fails Collectively: Robust Alignment Against Harmful Fine-Tuning via Joint Weight Perturbations▲5.4k3.00
[PDF]Refusals That Bend: Measuring and Predicting Task Malleability in Embodied VLM Planners▲3952.67
[PDF]HIPO: Heterogeneous and Instantaneous Policy Optimization for Safe Language Model Alignment▼4.3k2.00
[PDF]Beyond Refuse-or-Comply: Fine-Grained Policy-to-Action Grounding in Safety-Aligned Language Models▼5.8k2.33
[PDF]Hijacking the Instruction Hierarchy in LLM Agents: From Semantic Injection to Persistent Compromise▼1.2k3.00
[PDF]To Cache or Not to Cache: Understanding and Mitigating Safety Incompatibility in Cross-Model KV Cache Reuse▲3.0k2.33
[PDF]Safe at One Loop, Risky at Another: Aligning Safety Across Recurrent Depths in Looped Language Models▲1.9k3.67
[PDF]MobileGrail: Learning Compact Guardrails for On-Device deployable Mobile GUI Agent Safety▲3.0k2.33
[PDF]From Dilution to Alignment: Understanding and Mitigating Long-Context Safety Degradation▼4.3k2.00
[PDF]Risk Exposure Does Not Mean Unsafe Action: Evidence-Dynamics Alignment for Agent Safety▼1361.33
Withdrawn or desk rejected (24)
[PDF]Safe Vision-Language Models via Sparse Autoencoder Feature-Guided Embedding SteeringWithdrawn1.75
[PDF]False Claims, Credible Images: A Red-Teaming Benchmark for Commercial Image GeneratorsWithdrawn2.00
[PDF]Safe Alone, Unsafe Together: Compositional Curvature Safety Alignment for Prompt RobustnessWithdrawn2.00