arXiv:2606. 11266v1 Announce Type: new Abstract: The cost signal that constrained-RL algorithms optimize against is almost always reactive: the simulator emits a non-zero cost only after a collision has begun, and the Lagrange multiplier of PPO-Lagrangian grows only after the episode budget has been exceeded.
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...
arXiv:2607. 09706v1 Announce Type: new Abstract: Language models turn a worded situation into a numeric plan, and the dominant pipelines (NL4Opt, OptiMUS, ORLM, OR-LLM-Agent) commit to a single objective and point-valued coefficients, then solve once.
arXiv:2509. 08521v2 Announce Type: replace-cross Abstract: FMT$^{*}$ plans efficiently in static worlds by expanding a cost-ordered wavefront and collision-checking lazily, but its single-pass unvisited rule cannot revise paths when obstacles change.
arXiv:2607. 17240v1 Announce Type: new Abstract: When does a committed intermediate stage in an LLM reasoning pipeline earn its cost?
arXiv:2606. 16999v1 Announce Type: cross Abstract: Frozen small code models ( =45.
The paper reports on a large‑scale verified search experiment using a 30B language model on a laptop, evaluating three operator packages—schematic notebooks, named obstacles, and behavioural repulsion—in a factorial design across nine construction problems. Results show that the full composition of operators closes the seed‑to‑record gap more effectively than any single component, increases construction‑hash diversity, and that memory plus repulsion consistently avoids collapse. A frontier proposer achieves similar gains in far fewer samples, but the search ultimately stalls near a plateau where the reference family is adopted and optimized only when provided as code.
arXiv:2607. 17136v1 Announce Type: cross Abstract: Agentic computer-use RL is reported in single runs, and those numbers mislead.
VLCP (Vision Language Control Policy) is a training‑free robot manipulation approach that keeps a vision‑language model (VLM) frozen and uses it to generate short Python control functions. Unlike traditional methods that retry a fixed policy, VLCP rewrites the control code every K steps based on multi‑view RGB, proprioceptive state, and state delta, allowing failures to be corrected within the same episode. In a 57‑task MuJoCo/RoboVerse benchmark, VLCP achieves 35.1% pooled success versus 3.5% for a single‑query baseline, with a 27.3% within‑episode recovery rate on failed grasps and efficient token usage.
The paper introduces HARNESSEVO, a method that decomposes a large language model’s harness into four independently evolvable components—role, task‑strategy, tool/format‑rules, and reflection/control. Experiments on ALFWorld show that overall success rates are similar to flat‑string evolution, but the reflection/control component alone accounts for most of the performance gains. The study also finds that evenly distributing optimization budget across all slots can be detrimental; concentrating resources on the high‑credit control slot recovers lost performance, while on WebShop all slots remain ineffective, suggesting task‑specific differences in harness value.
SiLR introduces a structure‑preserving admission and process reward mechanism for large language model (LLM) tool agents. Unlike traditional scalar‑score gates that can trap agents in plateau trajectories, SiLR shadow‑executes each proposal and admits it based on a product order over branch‑level violation states, ensuring safe and recoverable actions. Experiments on Gym‑ANM and CityLearn benchmarks show SiLR consistently recovers all multi‑action episodes and outperforms scalar gates, while also providing a robust reward signal for policy learning.
arXiv:2606. 14238v1 Announce Type: cross Abstract: Safety certification of Vision-Language-Action (VLA) driving planners under ISO 21448 (SOTIF) rests on an Operational Design Domain (ODD) specification that answers two complementary questions: when does the planner start to fail, and how severely does it fail once it does?
Reinforcement learning from verifiable rewards (RLVR) for mathematical reasoning suffers from a structural blind spot: on "cliff" prompts-those on which every sampled rollout in a group fails-the group-normalized advantage is identically zero, so GRPO produces no gradient on precisely the prompts at the frontier of the model's capability. We introduce LoRA Scaffolded Policy Optimization (LSPO), a sampling-time mechanism that recovers this lost gradient.