AI Paper Atlas

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

NL2CA: Auto-formalizing Cognitive Decision-Making from Natural Language Using an Unsupervised CriticNL2LTL Framework

Zihao Deng, Yijia Li, Renrui Zhang, Peijun Ye

AAAI 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 2 citations (OpenAlex)

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 Critic Tree refinement on top of fine-tuned LLMfine-tuned Qwen3-0.6B · NL-to-LTL Dataset1accuracy: 22.2 -> 80.6Improved
add Critic Tree refinement on top of fine-tuned LLMfine-tuned Qwen3-0.6B · NL-to-LTL Dataset2accuracy: 97.5 -> 90.2Hurt
add Self-Refine to end-to-end DeepSeek R1DeepSeek R1 · NL-to-LTL Dataset1accuracy: 63.9 -> 63.9No clear effect
add Self-Refine to end-to-end DeepSeek R1DeepSeek R1 · NL-to-LTL Dataset1BLEU: 0.546 -> 0.490Hurt
use heterogeneous critic Tree instead of Self-RefineDeepSeek R1 · NL-to-LTL Dataset1accuracy: 63.9 -> 80.6Improved
interactive human-in-loop GPT-4 vs automated Critic TreeCriticNL2LTL · NL-to-LTL Dataset2accuracy: 78.3 -> 90.2Improved
let the LLM infer missing environment information when building rules (Msup) (ablation)Mlit · Highway driving simulationsuccess rate: 58.70 -> 12.00Hurt
let the LLM infer missing information (Msup) (ablation)Mlit · Lane Changing Interferencesuccess rate: 91.30 -> 63.82Hurt
use NL2CA literal rules (Mlit) instead of manual expert rulesmanual baseline · Driving simulation, average over 3 scenariossuccess rate: 65.30 -> 78.57Improved
use NL2CA literal rules instead of manual rulesmanual baseline · Driving simulationRSR: 99.47 -> 95.69Hurt
add cognitive reinforcement learning on human dataMlit vs manual · Driving simulationJS divergenceImproved

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