[Bug] Model generates incorrect "No valid results" summary, overwriting successful test run output
Status Fixed / completed
Reported on v2.1.185
Maintainer reply None cached
Activity 3 comments · opened Jun 21, 2026 · closed Jun 21, 2026
Bug Description
Model wrote false "No valid results" summary that erased its own graded 19/25 run, then admitted it when shown disk evidence. Same session ~7 min apart. https://claude.ai/code/session_01Vu3duGN1DBhnBzakF66eNt — full writeup in REPORT.md.
Environment Info
- Platform: darwin
- Terminal: ghostty
- Version: 2.1.185
- Feedback ID: 23999f73-6438-488f-bb6a-6584a1ca9ce0
Errors
[]
3 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Duplicate of #68990 (same root cause: model-generated text treated as ground truth → fabricated state). Full evidence + the inverse-polarity variant (fabricated failure erasing a real graded run, wrapped in a candor performance) posted there: https://github.com/anthropics/claude-code/issues/68990#issuecomment-4762699519
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.