cohesive-llm-benchmark · explorer

corpus single-turn LLM eval (200) multi-turn LLM eval (159 convs · 330 turns) multi-sample workflows LLM eval (5) single-turn LLM eval (50) multi-turn LLM eval (34 turns) category all error category all outcome all PASS only FAIL only tag all literal-match extras-best-practice extras-irrelevant missing-steps hallucinated upstream-rate-limited search