arXiv:2609.01274v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning, but how these gains relate to inference-time decoding and sear...
The paper investigates why reinforcement learning with verifiable rewards (RLVR) reduces the diversity of solutions in reasoning tasks. By analyzing the Countdown task, the authors show that RLVR contracts the solution space mainly at the entrance—before the first arithmetic operation—causing a 67% drop in solution coverage. They demonstrate that providing an unselected entrance prefix or applying entrance‑targeted interventions can restore or even improve coverage without harming accuracy.
Software engineering (SWE) agents resolve repository-level issues through long trajectories that grow increasingly expensive as context accumulates. Failed runs tend to be longer and exhibit redundant exploration or looping, suggesting that some failures may be detectable before completion.
arXiv:2608. 03222v1 Announce Type: cross Abstract: Software engineering (SWE) agents resolve repository-level issues through long trajectories that grow increasingly expensive as context accumulates.
The paper proposes CANOPY, a minimalist reinforcement learning protocol that addresses two common pitfalls—signal starvation and policy drift—in outcome‑only RL for long‑horizon interactive tasks. By scaling same‑task exploration, keeping updates on‑policy, and anchoring updates with KL divergence, CANOPY enables a Qwen3‑14B agent to achieve top leaderboard results on the AppWorld coding benchmark without auxiliary supervision or elaborate scaffolding. The approach also improves performance on SWE‑bench for a Qwen3.5‑9B model.
The paper introduces the concept of intervention fidelity in latent world models, measuring whether a model’s open‑loop transitions align with actual environment interventions. Experiments on TD‑MPC2, Cheetah, and DreamerV3 show that high reward fit does not guarantee fidelity, and that self‑supervised models can outperform task‑anchored ones in preserving intervention effects. The authors propose a capture‑gated audit to localize failures and argue that fidelity must be directly audited on the model’s native interface.
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."
The paper introduces HALTER, a graph-based system that automates the reset and evaluation of long-horizon robot manipulation tasks. HALTER constructs a spatial scene graph from point clouds and vision models, uses an LLM to score rollouts, plan resets, and verify success, all without labeled success images. In experiments on a Franka arm, HALTER restores scenes in 76% of episodes, improves skill completion estimation, and reduces operator time by 72% compared to manual reset.
arXiv:2608. 16889v1 Announce Type: cross Abstract: Long-horizon robot manipulation chains many contact-rich skills into one multi-stage task.
PROOF-Gen is a method that improves distillation of tool‑calling models by recovering successful trajectories from teacher failures. It uses per‑scenario prompt optimization to generate corrective guidance that steers the teacher to a passing trajectory, then removes this guidance before training so the student learns from clean demonstrations. On τ2‑bench, PROOF-Gen recovers 93% of failed scenarios, boosting Qwen3‑4B‑Instruct‑2507’s Pass^1 from 0.132 to 0.529 and improving Gemma 4 E4B‑it by 7.2pp on BFCL v4 multi‑turn, while also raising deployed on‑device model performance by up to 5.0pp across response‑quality metrics.
arXiv:2608. 06811v1 Announce Type: cross Abstract: Resolving a real software issue with a large language model (LLM) agent is a long repair episode, often tens to hundreds of steps spanning exploration, hypothesis, implementation, and verification.
arXiv:2608. 03483v1 Announce Type: cross Abstract: Existing chunk-based Vision-Language-Action (VLA) models execute a fixed number of actions (i.