arXiv Machine Learning June 11, 2026 By Samuel Tetteh, Cody Fleming

Seeing Before Colliding: Anticipatory Safe RL with Frozen Vision-Language Models

Read the original on arXiv Machine Learning →

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.

Summary generated by The Flow from the publisher's feed. The full article lives at arXiv Machine Learning.

arXiv AI

Jul 14

YUKTI: From Natural-Language Situations to Robust, Verifiable Decisions An Uncertainty-Typed Proposition IR, Assumption-Robust Pareto Frontiers, and a Regret Certificate

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.

By Suyash Mishra

llms agents benchmarks

arXiv AI

Jul 21

Constrained Path Reasoning: Measuring When Committed Stages Earn Their Cost

arXiv:2607. 17240v1 Announce Type: new Abstract: When does a committed intermediate stage in an LLM reasoning pipeline earn its cost?

By Honglin Li (ShanghaiTech University)

llms

arXiv Machine Learning

Jun 16

Selection Without Signal, Recovery Through Expression: A Measurement Study of Post-Hoc Falsification Operators for Frozen Small Code Models

arXiv:2606. 16999v1 Announce Type: cross Abstract: Frozen small code models ( =45.

By Mehmet Iscan

benchmarks safety

arXiv AI

Jul 21

Teach it to stop, not just to click

arXiv:2607. 17136v1 Announce Type: cross Abstract: Agentic computer-use RL is reported in single runs, and those numbers mislead.

By Barada Sahu (Cabal AI), Shivesh Pandey (Para AI)

agents reinforcement-learning efficiency multimodal

arXiv AI

Jun 15

When and How Severely: Scenario-Specific Safety Envelopes for Driving VLAs

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?

By Abhinaw Priyadershi, Jelena Frtunikj

multimodal safety

Hugging Face Trending Papers

Jul 30

LoRA Scaffolded Policy Optimization (LSPO): A Sampling-Time Low-Rank Scaffold for Recovering Reinforcement-Learning Gradient on Zero-Reward Cliff Prompts

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.

reinforcement-learning fine-tuning benchmarks

Related stories

YUKTI: From Natural-Language Situations to Robust, Verifiable Decisions An Uncertainty-Typed Proposition IR, Assumption-Robust Pareto Frontiers, and a Regret Certificate

Constrained Path Reasoning: Measuring When Committed Stages Earn Their Cost

Selection Without Signal, Recovery Through Expression: A Measurement Study of Post-Hoc Falsification Operators for Frozen Small Code Models

Teach it to stop, not just to click

When and How Severely: Scenario-Specific Safety Envelopes for Driving VLAs

LoRA Scaffolded Policy Optimization (LSPO): A Sampling-Time Low-Rank Scaffold for Recovering Reinforcement-Learning Gradient on Zero-Reward Cliff Prompts