AfterDeleteBench finds LLM agents can recover deleted secrets
One query recovered a deleted secret from Qwen2.5-7B-Instruct in about two thirds of cases, while never-exposed controls recovered none.
52
- Mentions · 7d0
- Mentions · 30d1
- All mentions2
- First seenSep 22, 2026
- Last mentioned20d ago
Timeline
2 articles · newest firstMeasuring the Serving Stack Instead of the Model: Hidden Confounds in Local Tool-Use Evaluation
Local serving stacks can distort tool-use scores, including false 0% fidelity for rejected models.
43
Appears with
- Mistralai model · 1
- SGLangproduct · 1
- Llama-3.2ai model · 1
- vLLMproduct · 1
- Qwen2.5-7B-Instructai model · 1
- llama.cppproduct · 1
- Ollamaproduct · 1
- Gemma-3ai model · 1
- AfterDeleteBenchproduct · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.