Claude mixes up who said what and that's not OK

Develop an automated evaluation suite that specifically tests multi-turn dialogue coherence regarding speaker attribution. This tool should identify instances where models conflate identities in complex chat logs to serve as a standard benchmarking dataset.

Suggested repo: speakerGuard

"Detect when your LLM forgets who said what before your users do."

Estimated effort: 40h