pi0 (AI model): news timeline & CVEs

Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models

One-word instruction edits swing VLA success sharply; rule-based rephrasing improves frozen policies without retraining.

The paper shows vision-language-action models remain highly sensitive to instruction wording even when built on vision-language models. Pi0.5 turns on a LIBERO stove 100% of the time for "switch on the stove" but only 2% for "switch on the hot plate," and a rephrase-augmented pi0 checkpoint still swings by up to 61 points. An oracle phrase search nearly closes a 21-point gap between in-distribution and out-of-distribution tasks. Distilling ten to twenty rephrasing rules and rewriting each instruction once improves frozen pi0 by 16–27% relative on twelve held-out tasks and lifts pi0.5 in-finetune LIBERO success from 93.6% to 97.8% without retraining.

arXiv cs.AI / cs.LG / cs.CLOct 7, 17:57 UTC · 1d agoAI research 3 sources1

Many Ways to Succeed: Diversity-Driven RL Fine-Tuning for VLA Generalization

DRIVE, a diversity-based RL fine-tune, lifts vision-language-action out-of-domain success, including +9.2 points on real dual-arm robots.

The paper introduces DRIVE, which turns diversity among successful behaviors into an intrinsic reward during reinforcement-learning fine-tuning of vision-language-action policies. Across LIBERO-Plus, ManiSkill3, and RoboTwin 2.0, average out-of-domain performance rises by 5.3 points on pi0 and 2.0 points on pi0.5 versus vanilla RL fine-tuning. On a dual-arm AgileX PiPER-X platform, average out-of-domain success increases from 64.1% to 73.3%.

arXiv cs.AI / cs.LG / cs.CLupdated Oct 7, 17:57 UTC · 1d agofirst Oct 7, 12:26 UTC · 1d agoAI research 3 sources1

Exploiting Vulnerabilities: Universal Adversarial Attacks on Vision-Language-Action Models in Robotics

A textured adversarial sphere cuts Pi0 and RDT robot task success by 31.2% to 39.9% in view.

Researchers introduce a Universal Adversarial Object, a sphere with an optimized surface texture, that degrades Vision-Language-Action robot policies when placed in view. A multi-level attack jointly disrupts trajectory planning, task execution, and action control. On Pi0 and RDT, average task success falls by 31.2% to 39.9% in simulation and real robots, and near zero in complex tasks.

arXiv cs.CRSep 30, 07:34 UTC · 8d agoAI safety & security

MemBodied: Recurrent Associative Memory for Vision-Language-Action Models

MemBodied gives vision-language-action policies fixed episodic memory without expanding context.

MemBodied adds a fixed-size episodic memory to vision-language-action policies so history-dependent manipulation does not require storing every past observation. It combines an associative state of interactions across policy calls with an episode anchor that keeps a compact view of the initial scene. On five memory-dependent RMBench tasks it achieves 7.81 times the mean success of a stateless policy and 2.98 times vanilla recurrent memory, beating the strongest memory baseline by 1.3 times with 10 times fewer added parameters. On fully observable LIBERO-Long it scores 90.6%, a 5.4-point gain over the stateless pi0 policy.

Hugging Face daily papersSep 22, 20:00 UTC · 16d agoAI research1

pi0

Timeline4 articles

Appears with

Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models

Many Ways to Succeed: Diversity-Driven RL Fine-Tuning for VLA Generalization

Exploiting Vulnerabilities: Universal Adversarial Attacks on Vision-Language-Action Models in Robotics

MemBodied: Recurrent Associative Memory for Vision-Language-Action Models