{"ID":2838983,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.16166","arxiv_id":"2511.16166","title":"EvoVLA: Self-Evolving Vision-Language-Action Model","abstract":"Long-horizon robotic manipulation remains challenging for Vision-Language-Action (VLA) models despite recent progress in zero-shot generalization and simulation-to-real-world transfer. Current VLA models suffer from stage hallucination, where agents exploit coarse evaluation signals to shortcut multi-step tasks, reporting high progress without truly completing them. We present EvoVLA, a self-supervised VLA framework that addresses this issue through three complementary components: Stage-Aligned Reward (SAR), which uses triplet contrastive learning with Gemini-generated hard negatives to prevent visual shortcuts; Pose-Based Object Exploration (POE), which grounds curiosity in relative object-gripper pose instead of raw pixels; and Long-Horizon Memory, which uses selective context retention and gated fusion to stabilize intrinsic shaping during extended rollouts. Extensive evaluations on Discoverse-L, a long-horizon manipulation benchmark with three multi-stage tasks, show that EvoVLA improves average task success by 10.2 percentage points over the strongest baseline (OpenVLA-OFT), reaching 69.2 percent. EvoVLA also achieves one-and-a-half times better sample efficiency and reduces stage hallucination from 38.5 percent to 14.8 percent. Real-world deployment on physical robots reaches an average success rate of 54.6 percent across four manipulation tasks, outperforming OpenVLA-OFT by 11 points, demonstrating effective sim-to-real transfer and strong generalization. Code: https://github.com/AIGeeksGroup/EvoVLA. Website: https://aigeeksgroup.github.io/EvoVLA.","short_abstract":"Long-horizon robotic manipulation remains challenging for Vision-Language-Action (VLA) models despite recent progress in zero-shot generalization and simulation-to-real-world transfer. Current VLA models suffer from stage hallucination, where agents exploit coarse evaluation signals to shortcut multi-step tasks, report...","url_abs":"https://arxiv.org/abs/2511.16166","url_pdf":"https://arxiv.org/pdf/2511.16166v1","authors":"[\"Zeting Liu\",\"Zida Yang\",\"Zeyu Zhang\",\"Hao Tang\"]","published":"2025-11-20T09:08:33Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"LoRA\"]","has_code":false,"code_links":[{"ID":606829,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2838983,"paper_url":"https://arxiv.org/abs/2511.16166","paper_title":"EvoVLA: Self-Evolving Vision-Language-Action Model","repo_url":"https://github.com/AIGeeksGroup/EvoVLA","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}