CONV-TO-BENCH: Evaluating Language Models via User–Assistant Dialogues in Code Tasks

Workshop Paper · April 2026 · 1 min read

CONV-TO-BENCH creates evaluation benchmarks from raw user–assistant interaction logs in code tasks, enabling model assessment on realistic multi-turn dialogue data.

undefined ⧉

CONV-TO-BENCH: Evaluating Language Models via User–Assistant Dialogues in Code Tasks

We present CONV-TO-BENCH, a framework for constructing evaluation benchmarks from raw user–assistant interaction logs in code-related tasks.