Long-term Traffic Simulation via Structured Autoregressive Modeling

arXiv 2026 41.6 method

TLDR

RosettaSim uses frozen LLMs for structured autoregressive traffic simulation, achieving state-of-the-art short- and long-term accuracy on Waymo.

Reasoning

The paper introduces a novel framework (RosettaSim) that leverages LLMs for traffic simulation, supported by probing experiments and a new evaluation method (RTE). Strengths include strong empirical results on a real-world benchmark and addressing dynamic token cardinality. Weaknesses include limited scope to traffic domain and reliance on frozen LLMs without exploring full fine-tuning.

Read-first score

Read-first score 41.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 43.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=evaluation,experiment,metric

Topical relevance 29%

61.4

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 281.

Keyword Scores

world model

world simulator

interactive world model

generative world model

world dynamics prediction

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Leveraging synergy between architectural inductive biases and statistical priors of LLMs for traffic simulation
Probing experiments showing transferability of attention mechanisms and distributional consistency between motion tokens and natural language, enabling small-scale frozen LLMs to adapt
RosettaSim: unified framework projecting scene topology, agent states, and spawning intents into a structured autoregressive stream with variable length
Retrieval-based Traffic Evaluation (RTE) that retrieves semantically similar real-world scenarios as context-aware reference anchors for long-horizon evaluation

Methodology

The paper proposes RosettaSim, a framework that adapts small-scale, heavily frozen LLMs by projecting scene topology, agent states, and spawning intents into a structured autoregressive stream with variable length. It also introduces RTE, which retrieves semantically similar real-world scenarios as reference anchors for evaluating extended rollouts. Experiments are conducted on the Waymo Open Sim Agent Challenge (WOSAC).

Key Results

RosettaSim achieves state-of-the-art performance in both short- and long-term simulation on WOSAC. RTE shows stronger correlation with standard metrics (r=0.83) than existing approaches (r=0.74).

Resources

Paper