AI Paper Atlas

Research map / Large language models / Agents, code and math

BIRD-INTERACT: Re-imagining Text-to-SQL Evaluation via Lens of Dynamic Interactions

Nan Huo, Xiaohan Xu, Jinyang Li, Per Jacobsson, Shipei Lin, Bowen Qin, Binyuan Hui, Xiaolong Li, Ge Qu, Shuzheng Si, Linheng Han, Edward Alexander and 12 more

ICLR 2026 · Large language models › Agents, code and math · approach: Software engineering agents

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 changedWhereMeasuredEffect
add function-driven two-stage mapping to the user simulatorGPT-4o user simulator · Alignment with human usersPearson correlation: 0.61 -> 0.84Improved
add function-driven two-stage mapping to the user simulatorGemini-2.0-Flash user simulator · Alignment with human usersPearson correlation: 0.54 -> 0.79Improved
add function-driven mapping to the user simulatoruser simulator · UserSim-Guard unanswerable questionsfailure rate: 67.4% -> 2.7%Improved
add debugging phase in c-InteractClaude-Sonnet-4 · BIRD-Interact-Full c-Interact prioritysuccess rate (%): +6.67Improved
add debugging phase in c-InteractGPT-5 · BIRD-Interact-Full c-Interact prioritysuccess rate (%): +0.67No clear effect
give GPT-5 the ambiguity-resolution histories of Qwen-3-Coder or O3-Mini (memory grafting)GPT-5 · BIRD-Interact c-Interactsuccess rateImproved
switch from c-Interact to a-Interact modeGPT-5 · BIRD-Interact-Full priority questionssuccess rate (%): 14.50 -> 29.17Improved
switch from c-Interact to a-Interact modeQwen-3-Coder-480B · BIRD-Interact-Full priority questionssuccess rate (%): 22.00 -> 13.33Hurt
switch from c-Interact to a-Interact modeClaude-Sonnet-4 · BIRD-Interact-Full priority questionssuccess rate (%): 22.33 -> 27.83Improved
increase user patience (more interaction turns)Claude-3.7-Sonnet · BIRD-Interact-LiteperformanceImproved
follow-up sub-tasks versus priority sub-tasksGPT-5 a-Interact · BIRD-Interact-Fullsuccess rate (%): 29.17 -> 17.00Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy