AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

TrafficAlign: Aligning Large Language Models for Traffic Scenario Generation

Zhi Tu, Liangkun Niu, Tianyi Zhang

CVPR 2026 · Large language models › Instruction and fine-tuning · approach: Calibration and knowledge in LLM tuning · 0 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 LLM alignment on video-derived scenarios (vs GPT-5 prompted directly) (ablation)TrafficAlign scenario generator · SafeBench testing, Los Angelescollision rate: 0.889 -> 0.933Improved
add LLM alignment (vs Claude Sonnet 4 prompted directly) (ablation)TrafficAlign scenario generator · SafeBench testing, Los Angelescollision rate: 0.894 -> 0.933Improved
add LLM alignment (vs Qwen3 prompted directly) (ablation)TrafficAlign scenario generator · SafeBench testing, Los Angelesoverall score: 0.450 -> 0.405Improved
use TrafficAlign scenarios instead of ChatScene for testingdriving models · SafeBench testing, Los Angelescollision rate: 0.825 -> 0.933Improved
use TrafficAlign scenarios instead of ChatScene for testingdriving models · SafeBench testing, Los Angelesoverall score: 0.481 -> 0.405Improved
use TrafficAlign scenarios instead of ChatScene for testingdriving models · SafeBench testing, Small Townsrunning red lights: 0.191 -> 0.186No clear effect
fine-tune driving models on TrafficAlign scenarios vs no fine-tuningPPO/SAC/TD3 driving models · Held-out set, Los Angelescollision rate: 0.968 -> 0.701Improved
fine-tune on TrafficAlign scenarios instead of ChatScene scenariosPPO/SAC/TD3 driving models · Held-out set, Los Angelescollision rate: 0.733 -> 0.701Improved
fine-tune on TrafficAlign scenarios instead of ChatScene scenariosPPO/SAC/TD3 driving models · Held-out set, Yosemitecollision rate: 0.641 -> 0.604Improved
fine-tune on TrafficAlign scenarios instead of ChatScene scenariosPPO/SAC/TD3 driving models · Held-out set, Los Angelesoverall score: 0.301 -> 0.385Improved
use 5s instead of 15s frame sampling interval on intense driving videos (ablation)TrafficAlign · SafeBench testing, 10 intense videoscollision rate: 0.899 -> 0.907Improved
use 5s instead of 15s frame sampling interval on intense driving videos (ablation)TrafficAlign · Fine-tuning driving models, 10 intense videoscollision rate: 0.768 -> 0.766No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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