OpenSafeIntent benchmark shows models fail to calibrate safety across intent shifts in matched dual-use prompts, indicating current evaluations are insufficient.
org/abs/2502.01042
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 6roles
background 1polarities
background 1representative citing papers
kNNGuard classifies prompts using multi-layer kNN on LLM hidden activations from 50 examples, matching or exceeding fine-tuned guardrails in F1 while running 2.7x to 10x faster with no training required.
Final-token probes miss distributed unsafe evidence in jailbreaks, but a PCA-HMM model on prefill trajectories recovers many misses without naive pooling's false positives.
Reasoning traces in large reasoning models expose safety failures missed by final-answer checks, and adaptive multi-principle steering reduces unsafe content in both traces and answers while preserving task performance.
GSAE improves selective refusal on safety benchmarks by smoothing SAE directions over a co-activation graph and applying them via a two-gate controller, outperforming standard SAEs and baselines on Llama-3 and other models.
Self-aligned reward uses relative perplexity differences to encourage concise, query-specific reasoning in LLMs, yielding 4% accuracy gains and 30% lower inference cost when added to PPO or GRPO.
citing papers explorer
-
OpenSafeIntent: Evaluating Intent-Calibrated Safe Completion Across Dual-Use Prompt Sets
OpenSafeIntent benchmark shows models fail to calibrate safety across intent shifts in matched dual-use prompts, indicating current evaluations are insufficient.
-
kNNGuard: Turning LLM Hidden Activations into a Training-Free Configurable Guardrail
kNNGuard classifies prompts using multi-layer kNN on LLM hidden activations from 50 examples, matching or exceeding fine-tuned guardrails in F1 while running 2.7x to 10x faster with no training required.
-
Before the Last Token: Diagnosing Final-Token Safety Probe Failures
Final-token probes miss distributed unsafe evidence in jailbreaks, but a PCA-HMM model on prefill trajectories recovers many misses without naive pooling's false positives.
-
Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering
Reasoning traces in large reasoning models expose safety failures missed by final-answer checks, and adaptive multi-principle steering reduces unsafe content in both traces and answers while preserving task performance.
-
Graph-Regularized Sparse Autoencoders for LLM Safety Steering
GSAE improves selective refusal on safety benchmarks by smoothing SAE directions over a co-activation graph and applying them via a two-gate controller, outperforming standard SAEs and baselines on Llama-3 and other models.
-
Self-Aligned Reward: Towards Effective and Efficient Reasoners
Self-aligned reward uses relative perplexity differences to encourage concise, query-specific reasoning in LLMs, yielding 4% accuracy gains and 30% lower inference cost when added to PPO or GRPO.