Evidence · DiggingBeagle record
CVV Dataset replication repository: README, taxonomy and model-output inventory
The browser-readable README describes audit-scoped source packages, normalized findings using 11 pattern IDs, two prompts and result directories for Composer-2.5, DeepSeek-V4-Pro, GPT-5.5 and Opus-4.8. The visible inventory lists 786 contract files and 836 result files. Individual entries were not exhaustively checked and evaluation scores were not recomputed.
- Source role
- primary disclosure
- Version
- Public README and visible inventory inspected 2026-10-13
Evidence record
The browser-readable README describes audit-scoped source packages, normalized findings using 11 pattern IDs, two prompts and result directories for Composer-2.5, DeepSeek-V4-Pro, GPT-5.5 and Opus-4.8. The visible inventory lists 786 contract files and 836 result files. Individual entries were not exhaustively checked and evaluation scores were not recomputed.
Claim-level citations (1)
- supportsBridge-auditing benchmark exposes uneven cross-component verification reasoning: The publicly readable replication README describes audit-scoped code packages, normalized 11-pattern finding labels and outputs for the four evaluated model families. It tells readers to retain missing dependencies as missing context.
README.md: findings/, prompts/ and results/; browser-visible repository inventory inspected 2026-10-13.