AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

InfGen: Scenario Generation as Next Token Group Prediction

Zhenghao Peng, Yuxin Liu, Bolei Zhou

ICLR 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Data-driven traffic simulation

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 autoregressive structured agent-state decoding (vs parallel independent heads) (ablation)SceneStreamer · WOMD initial state MMD strict, SizeMMD (lower better): 0.1172 -> 0.0743Improved
add autoregressive agent-state decoding (ablation)SceneStreamer · WOMD initial state MMD strict, PositionMMD (lower better): 0.1603 -> 0.1291Improved
add autoregressive agent-state decoding (ablation)SceneStreamer · WOMD initial state MMD strict, HeadingMMD (lower better): 0.1646 -> 0.1270Improved
add autoregressive agent-state decoding (ablation)SceneStreamer · WOMD initial state MMD strict, VelocityMMD (lower better): 0.2114 -> 0.1970Improved
add autoregressive agent-state decoding (ablation)SceneStreamer · WOMD initial state MMD relaxed (all agents), PositionMMD (lower better): 0.3237 -> 0.2198Improved
add autoregressive agent-state decoding (ablation)SceneStreamer · WOMD initial state MMD relaxed (all agents), VelocityMMD (lower better): 0.1183 -> 0.0730Improved
train planner on SceneStreamer-Motion traffic instead of log-replayTD3 RL planner · WOMD val (100 scenarios)Average episodic reward: 32.24 -> 37.98Improved
add adaptive SDC state-forcing (closed-loop training) (ablation)SceneStreamer-Motion planner training · WOMD val (100 scenarios)Average episodic reward: 37.98 -> 39.23Improved
add adaptive SDC state-forcing (ablation)SceneStreamer-Full (No RS) · WOMD val (100 scenarios)Average episodic reward: 38.18 -> 38.81No clear effect
add reject sampling for colliding agents (ablation)SceneStreamer-Full (w/ Ada) · WOMD val (100 scenarios)Success rate: 0.7385 -> 0.7620Improved
add reject sampling for colliding agents (ablation)SceneStreamer-Full (w/ Ada) · WOMD val (100 scenarios)Route completion: 0.7230 -> 0.7345Improved
fine-tune to also generate agent states (SceneStreamer-Full vs Motion)SceneStreamer · WOMD motion prediction (all agents)ADEmin: 0.8730 -> 0.9339Hurt

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