cd /news/large-language-models/llm-safety-why-asr-is-a-bad-metric-f… · home topics large-language-models article
[ARTICLE · art-71032] src=promptcube3.com ↗ pub= topic=large-language-models verified=true sentiment=· neutral

LLM Safety: Why ASR is a Bad Metric for Defenders

A new framework called A-MESS, which uses AttackSHAP Shapley-based scoring, shows that Attack Success Rate (ASR) is a poor metric for improving LLM safety because high ASR correlates weakly with actual safety utility. The findings demonstrate that selecting a compact subset of high-utility attacks is significantly more effective than including all successful attacks in training, making the approach practical for real-world LLM agent development.

read1 min views1 publishedJul 23, 2026
LLM Safety: Why ASR is a Bad Metric for Defenders
Image: Promptcube3 (auto-discovered)

everysuccessful attack; you need the ones that actually move the needle on model resilience.

This is where a defender-centric approach changes the game. Instead of asking "Did this break the model?", the question becomes "Does including this attack in my training set actually improve downstream safety?"

The A-MESS framework tackles this by using AttackSHAP—a Shapley-based scoring system. For those unfamiliar, Shapley values are typically used in game theory to determine how much each player contributes to a total outcome. In this context, it treats individual jailbreak attacks as "players" and calculates their marginal utility toward improving the model's safety.

The findings here are pretty revealing:

ASR vs. Utility: High attack success rates correlate weakly with actual safety utility. A "noisy" jailbreak might have a high ASR but provide zero value for alignment training.Subset Optimization: Selecting a compact subset of high-utility attacks is significantly more effective than just dumping every successful attack into the training pipeline.Efficiency: AttackSHAP can be estimated accurately without needing an infinite number of utility queries, making it practical for real-world LLM agent development.

Essentially, we need to stop treating jailbreaks as just "failures" to be patched and start treating them as high-value data resources. The goal isn't just to stop one specific prompt, but to identify the underlying vulnerability that the prompt exposed.

If you're building a safety pipeline from scratch, focusing on marginal utility rather than raw success rates will likely lead to a much leaner and more effective red-teaming dataset.

[Next Preemptive Hardening for Agentic LLM Security →](/en/threads/2326/)
── more in #large-language-models 4 stories · sorted by recency
── more on @a-mess 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/llm-safety-why-asr-i…] indexed:0 read:1min 2026-07-23 ·