Research map / Large language models / Agents, code and math
Talk2Code: A Multi-Turn Interaction Benchmark with Dual-Track Evaluation for Code Generation
Weibin Yang, Liangru Xie, Jieyun Cai, Yuxiang Yan, Hong-Ning Dai, Hao Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| switch from direct generation with golden context to interactive clarification dialogue | GPT-4o · Talk2Code (Novice users) | Test Pass Rate / Success Rate: 0.54/0.43 -> 0.41/0.30 | Hurt |
| switch from direct generation with golden context to interactive clarification dialogue | o1-mini · Talk2Code (Novice users) | Test Pass Rate / Success Rate: 0.59/0.47 -> 0.38/0.27 | Hurt |
| switch from direct generation to interactive dialogue | GLM-4-9B · Talk2Code (Novice users) | Test Pass Rate: 0.43 -> 0.16 | Hurt |
| move from novice to intermediate users | DeepSeek-R1 · Talk2Code Direct Generation | Test Pass Rate: 0.57 -> 0.63 | Improved |
| move from intermediate to expert users | DeepSeek-R1 · Talk2Code Direct Generation | Test Pass Rate: 0.63 -> 0.59 | Hurt |
| move from intermediate to expert users | LLaMA-3.3-70B · Talk2Code Interactive Dialogue | subjective average: 3.42 -> 2.77 | Hurt |
| move from intermediate to expert users | GPT-4o · Talk2Code Interactive Dialogue | subjective average: 3.78 -> 3.66 | Hurt |
| move from novice to expert users in interactive dialogue | DeepSeek-R1 · Talk2Code Interactive Dialogue | Test Pass Rate: 0.44 -> 0.49 | Improved |
| user expertise level (novice, intermediate, expert) | GPT-4o · Talk2Code Direct Generation | Test Pass Rate: 0.54 / 0.61 / 0.57 | Mixed |
Closest papers by meaning
- ConvCodeWorld: Benchmarking Conversational Code Generation in Reproducible Feedback Environments (ICLR 2025 · similarity 0.90)
- ToolDial: Multi-turn Dialogue Generation Method for Tool-Augmented Language Models (ICLR 2025 · similarity 0.88)
- MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback (ICLR 2024 · similarity 0.88)
- Talking Turns: Benchmarking Audio Foundation Models on Turn-Taking Dynamics (ICLR 2025 · similarity 0.86)
- ToolACE-MT: Non-Autoregressive Generation for Agentic Multi-Turn Interaction (ICLR 2026 · similarity 0.86)
- CollabLLM: From Passive Responders to Active Collaborators (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
