Hi-FLoop introduces a hierarchical state‑feedback framework for multi‑agent traffic simulation that reconciles decision time scales over an 8‑second rollout. The model uses eight scene‑level Worlds to maintain joint hypotheses, with an 8‑second Goal, 2‑second Preview, and 1‑second Control hierarchy, and commits only executed prefixes every 0.5 seconds to preserve factual consistency. A joint preview interaction graph and a prefix‑frozen A‑to‑B cascade enable sparse interaction refinement and accurate state recovery, achieving an overall score of 0.689987 on the H‑D public‑validation split and strong oracle‑minADE performance. whyItMatters":"The paper presents a novel multi‑timescale approach that improves consistency and realism in long‑horizon traffic simulations, as evidenced by its competitive evaluation metrics."
arXiv:2609.08796v2 Announce Type: cross Abstract: Multi-agent traffic simulation seeks diverse, coordinated, and physically realistic futures from maps and observed history. Long-horizon closed-loop...
Agentic World Modeling (AWM) redefines next-observation prediction as a multi-turn evidence-acquisition task, allowing agents to query a real environment safely and construct observations without mutating shared state. Integrated with trajectory-level test-time scaling, AWM enables parallel exploration over a single shared environment, outperforming single-shot simulation by 40.07% and trained world-model baselines by 12.36% on benchmarks such as τ-bench, AppWorld, and OfficeBench. This approach provides a practical foundation for safe and scalable agentic test-time scaling in shared, non-resettable environments.
ActionSplice is an inference framework for chunk‑autoregressive video world models that allows in‑flight action editing without re‑sampling completed evaluations. It formulates the problem as Counterfactual State Transport (CST), using a lightweight corrector to move the backbone representation toward the state induced by a revised action at the same solver step. Two variants, CST*R and CST*T, update either the entire active chunk or only its suffix, achieving significant reductions in rollback‑relative LPIPS and providing speedups over waiting.
The paper introduces SLIM, a pushing benchmark that tests language‑based goal specification for latent world models. It shows that a vanilla LeWM fails almost entirely due to an action‑insensitive encoder, but adding an inverse‑dynamics auxiliary loss restores encoder sensitivity and dramatically improves success rates. With the repaired latent, a lightweight language‑goal head can plan from natural language sentences, achieving high performance on navigation and pushing tasks without retraining the world model.
An interactive world model must follow the user's actions, remember the places it has shown, and stream in real time. The tension is structural: control wants a short horizon, memory wants an unbounde...
The paper introduces a new evaluation protocol called checkpoint handoff to disentangle the contributions of reaching a target state and solving the task in reinforcement learning agents. By cloning states reached by one checkpoint and handing them to another without retraining, the authors separate the REACH metric (how often a policy arrives at a state confirmed to be a fixed number of actions from success) from the SOLVE metric (how often it finishes from that identical state). Across two benchmarks and pipelines, the analysis shows that RL history benefits RL solvers more than SFT solvers, and that independent REACH and SOLVE gaps predict overall performance.
The paper introduces a denoised counterfactual protocol to accurately measure the selectable headroom of world action models (WAMs) and uses these measurements to build a fail‑gated selector. Across four WAMs and two benchmarks, the study shows that the standard estimate overstates selectable headroom, that structured action perturbations can add up to 16 points of headroom, and that imagined futures retain candidate‑outcome information without loss. The selector improves closed‑loop task success from 76.9 % to 81.6 % on held‑out training episodes and from 62.6 % to 69.8 % on unseen tasks.
The paper introduces CoPE, a method that trains language agents by adding multi‑step action targets to the policy backbone. By supervising each action forecasted from preceding histories, CoPE enables acting and forecasting to co‑evolve without extra inference cost. Across five benchmarks, CoPE outperforms existing RL baselines, achieving up to twice the sample efficiency on ScienceWorld and improving overall performance on realistic applications.
FIRM-WM is a compact pixel world model that separates a goal‑comparable configuration from a 128‑dimensional dynamic fiber, enabling reward‑free visual planning from offline videos. It addresses two key mismatches: aligning planning states with goal images and reconciling factual trajectories with interventional sampling. In experiments, FIRM‑WM achieves high success rates on TwoRoom, Reacher, and OGBench‑Cube while using fewer parameters and faster planning times than prior models.
The paper challenges the common practice of selecting latent world models based on one‑step prediction loss, arguing that this metric poorly predicts planning performance. Using a frozen‑decoder protocol, the authors decompose one‑step error into encoder‑inherent and predictor‑added components and evaluate goal‑conditioned planning across three continuous‑control environments. Their experiments show that improvements in one‑step fidelity do not correlate with planning success, with correlation coefficients near zero and inconsistent exchange rates across tasks.
The paper introduces TSBPO, a tree‑structured sampling‑to‑update framework for training interactive language agents. It tackles three intertwined challenges—sampling task and state continuations, assigning delayed feedback, and aggregating token losses—by using a likelihood‑fitted Root/Branch allocator, trajectory and observation‑group advantages, and a clipped token surrogate aggregation. Experiments with a Qwen2.5‑1.5B actor show high success rates on ALFWorld and WebShop benchmarks, demonstrating the effectiveness of the proposed training configuration.
A ReAct-style agent accumulates observations, intermediate results, and failed attempts in a single context. As this context grows, stale and off-path information consumes its limited capacity and obs...