Output-Level Regularization Eliminates the Seed Lottery in Single-GPU VLA Fine-Tuning

[2606.13856] Output-Level Regularization Eliminates the Seed Lottery in Single-GPU VLA Fine-Tuning Skip to main content Search Submit Donate Log in Computer Science > Robotics arXiv:2606.13856 (cs) This paper has been withdrawn by Jeffrin Sam [Submitted on 11 Jun 2026 ( v1 ), last revised 25 Aug 2026 (this version, v2)] Title: Output-Level Regularization Eliminates the Seed Lottery in Single-GPU VLA Fine-Tuning Authors: Jeffrin Sam , Dzmitry Tsetserukou No PDF available, click to view other formats Abstract: Fine-tuning a vision-language-action model (VLA-JEPA) on a single GPU should be simple: load a pretrained checkpoint, run training, deploy. There is a hidden danger. Run the same fine-tuning code thirteen times -- same data, same architecture, different random seed -- and twelve runs produce a robot succeeding 91--94% of the time, while one run silently degrades to 65.2%: a 29 pp gap with no error message, no warning, and no way to predict which seed will fail. We call this the seed lottery. We trace the cause to output collapse: the action predictor quietly learns to produce nearly identical outputs regardless of what the robot sees. Existing weight-level methods (L2, EWC) are structurally blind to this collapse -- they penalize weight changes, but collapse occurs in directions weights can move freely without affecting outputs, a gap we formalize via the Jacobian null-space. Across 7 methods x up to 13 seeds x 3 LIBERO benchmarks, three output-level regularizers -- VICReg (n=12 seeds), Dropout (n=4), and a halved learning rate (n=5) -- each eliminate every catastrophic seed (0/21 combined collapses vs. 1/13 Baseline; F(12,11)=28.7, p<0.001), while weight-level methods (L2, EWC) preserve the lottery. The simplest fix is changing one number in your optimizer config. Comments: Some errors in the reference, and Im still developing to make it more novel Subjects: Robotics (cs.RO) ACM classes: I.2.9; I.2.6; I.5.1 Cite as: arXiv:2606.13856 [cs.RO] (or arXiv:2606.13856v2 [cs.RO] for this version) https://doi.org/10.48550/arXiv.2606.13856 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Jeffrin Sam [ view email ] [v1] Thu, 11 Jun 2026 19:33:11 UTC (399 KB) [v2] Tue, 25 Aug 2026 06:54:20 UTC (1 KB) (withdrawn) Full-text links: Access Paper: Withdrawn Additional Features Audio Summary Current browse context: cs.RO < prev | next > new | recent | 2026-06 Change to browse by: cs References & Citations SciX Google Scholar Semantic Scholar export BibTeX citation Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? ) Connected Papers Toggle Connected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) scite.ai Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle Gotit.pub ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle TXYZ.AI ( What is TXYZ.AI? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs . Which authors of this paper are endorsers? | Disable MathJax ( What is MathJax? ) We gratefully acknowledge support from our major funders , member institutions , and all contributors. About · Help · Contact · Subscribe · Copyright · Privacy · Accessibility · Operational Status (opens in new tab)

Related documents

• Output-Level Regularization Eliminates the Seed Lottery in Single-GPU VLA Fine-Tuning#670633
• NUNA: Characterizing and Mitigating Non-Uniform Network Access in Multi-Die GPU Scale-Up Systems#423941
• FliX: Flipped-Indexing for Scalable GPU Queries and Updates#120487
• Reshaping Rollout Workloads for Asynchronous RL Post-Training on Heterogeneous Accelerators#1132100
• Reshaping Rollout Workloads for Asynchronous RL Post-Training on Heterogeneous Accelerators#1122264