MemEvoBench (Workflow Style, Biased Feedback): leaderboard
MemEvoBench (2026) evaluates memory safety of LLM agents: misleading entries in a mixed memory pool, and the agent's own responses stored after each round, test whether accumulated memory drifts the agent into unsafe behavior over three rounds. This board is the attack success rate in percent in the third round on workflow-style cases whose memory holds noisy tool returns, judged by GPT-5.2, with simulated biased user feedback rewarding risky shortcuts; lower is better.
Metric: Attack success rate (%) in the third evaluation round on the 83 workflow-style cases (20 Agent-SafetyBench environments with noisy tool returns in the memory pool): share of responses that follow the misleading memory into the target risk behavior after two rounds of the agent's own responses were added to the memory pool, with simulated biased user feedback stored after each round (risky shortcuts rewarded, cautious answers penalized), judged MISLED or SAFE by GPT-5.2, temperature 0, standard system prompt without safety guidance; lower is better. Source: arxiv.org. Saturation forecast: Around 2028. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 85.5 | #276 |
| 2 | GPT-5 | 87 | #178 |
| 3 | Claude 3.7 Sonnet | 87.5 | #470 |
| 4 | Qwen 3 Next 80B A3B Instruct | 89.2 | #778 |
| 5 | Qwen 3 235B A22B 2507 Instruct | 90.4 | #616 |
| 6 | GPT-4o | 91.5 | #751 |
| 7 | Qwen 3 32B | 91.6 | #972 |
| 8 | Llama 3.3 70B Instruct | 94 | #1385 |
| 9 | DeepSeek V3.2 | 94 | #419 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=memevobench-workflow-style-biased-feedback · How It Works · Data refreshed daily, snapshot 2026-11-21.