Primitive Subspaces Mediate Few-Shot Transfer in VLAs

• URL: http://arxiv.org/abs/2605.30695v1
• Date: Fri, 29 May 2026 00:37:53 GMT
• Title: Primitive Subspaces Mediate Few-Shot Transfer in VLAs
• Abstract summary: We investigate whether primitive-aware training produces a transferable artifact.<n>We train two VLA architectures with different inductive biases.<n>We show few-shot transfer degrades by 32 percentage points while ablating a random subspace of equal dimensionality has no effect.
• Score: 1.3918848543076061
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Deploying vision-language-action (VLA) policies in industrial environments requires the ability to teach new tasks at low cost, a property current VLAs lack, since each new task requires fine-tuning. We investigate whether primitive-aware training produces a transferable artifact: a learned library of sub-skills that can be composed at inference time, conditioned on a small number of demonstrations, to perform tasks the policy was never trained on. We train two VLA architectures with different inductive biases, OpenVLA and $π_{0.5}$, on the REASSEMBLE contact-rich assembly dataset under matched LoRA fine-tuning recipes and locked hyperparameters, varying training between flat trajectories and primitive-segmented episodes with primitive-specific language prompts. We hold out 6 object-task combinations from training and evaluate few-shot transfer: models receive $m \in \{0, 1, 3, 5, 10\}$ demonstrations of a held-out task and attempt execution without weight updates. We replicate across three training seeds and validate on a second dataset (LIBERO-Long). Primitive-trained models reach 78% of fine-tuned upper-bound performance with only m=3 demonstrations, while flat-trained models require m=10 demonstrations to reach the same level -- a $3\times$ sample efficiency gap that replicates across seeds, architectures, and datasets. To establish causation, we ablate the primitive-decodable subspace of hidden states and show few-shot transfer degrades by 32 percentage points while ablating a random subspace of equal dimensionality has no effect, indicating primitive representations are causally necessary rather than incidentally correlated with transfer. We identify and correct a methodological pitfall in evaluating chunked policies: family-wise inflation of single-step action-range gates produces order-of-magnitude higher false-failure rates against ground-truth human demonstrations.

Related papers

• Learning to Move Before Learning to Do: Task-Agnostic pretraining for VLAs [46.61554174146852]
  Task-Agnostic Pretraining (TAP) is a two-stage framework that first learns transferable motor priors from cheap, unlabeled interaction data.<n>On the SIMPLER benchmark, TAP matches models trained on over 1M expert trajectories while using orders of magnitude less labeled data.<n>On a real-world WidowX platform, TAP retains 25% success under camera perturbations where internet-scale baselines collapse to 0%.
  arXiv  Detail & Related papers  (2026-07-02T17:33:37Z)
• VISTA: Vision-Grounded and Physics-Validated Adaptation of UMI data for VLA Training [52.05483137072975]
  Universal Manipulation Interface (UMI) enables scalable real-world robot data collection without hardware-specific teleoperation.<n>We present VISTA, a framework that bridges this dual gap through three synergistic components.<n>We release the physical-validation pipeline, UMI-VQA, validated trajectory data, and the pre-trained model for the community.
  arXiv  Detail & Related papers  (2026-06-03T10:38:45Z)
• Universal Pose Pretraining for Generalizable Vision-Language-Action Policies [83.39008378156647]
  Existing Vision-Language-Action (VLA) models often suffer from feature collapse and low training efficiency.<n>We propose Pose-VLA, a decoupled paradigm that separates VLA training into a pre-training phase for extracting universal 3D spatial priors.<n>Our framework follows a two-stage pre-training pipeline, establishing fundamental spatial grounding via poses followed by motion alignment.
  arXiv  Detail & Related papers  (2026-02-23T11:00:08Z)
• Bagging-Based Model Merging for Robust General Text Embeddings [73.51674133699196]
  General-purpose text embedding models underpin a wide range of NLP and information retrieval applications.<n>We present a systematic study of multi-task training for text embeddings from two perspectives: data scheduling and model merging.<n>We propose Bagging-based rObust mOdel Merging (BOOM), which trains multiple embedding models on sampled subsets and merges them into a single model.
  arXiv  Detail & Related papers  (2026-02-05T15:45:08Z)
• Steering Vision-Language-Action Models as Anti-Exploration: A Test-Time Scaling Approach [78.4812458793128]
  We propose textbfTACO, a test-time-scaling framework that applies a lightweight pseudo-count estimator as a high-fidelity verifier of action chunks.<n>Our method resembles the classical anti-exploration principle in offline reinforcement learning (RL), and being gradient-free, it incurs significant computational benefits.
  arXiv  Detail & Related papers  (2025-12-02T14:42:54Z)
• Basis-Oriented Low-rank Transfer for Few-Shot and Test-Time Adaptation [10.804106052326402]
  Adapting large pre-trained models to unseen tasks under tight data and compute budgets remains challenging.<n>We propose BOLT, a framework that reuses existing fine-tuned models and adapts within that subspace.<n>Our results show that constraining adaptation to a task-informed subspace provides an effective alternative for unseen-task transfer.
  arXiv  Detail & Related papers  (2025-12-02T06:00:16Z)
• Detect Anything via Next Point Prediction [51.55967987350882]
  Rex- Omni is a 3B-scale MLLM that achieves state-of-the-art object perception performance.<n>On benchmarks like COCO and LVIS, Rex- Omni attains performance comparable to or exceeding regression-based models.
  arXiv  Detail & Related papers  (2025-10-14T17:59:54Z)
• FINE: Factorizing Knowledge for Initialization of Variable-sized Diffusion Models [35.40065954148091]
  FINE is a method based on the Learngene framework to initializing downstream networks leveraging pre-trained models. It decomposes pre-trained knowledge into the product of matrices (i.e., $U$, $Sigma$, and $V$), where $U$ and $V$ are shared across network blocks as learngenes'' It consistently outperforms direct pre-training, particularly for smaller models, achieving state-of-the-art results across variable model sizes.
  arXiv  Detail & Related papers  (2024-09-28T08:57:17Z)
• Continual learning with task specialist [2.8830182365988923]
  We propose Continual Learning with Task Specialists (CLTS) to address the issues of catastrophic forgetting and limited labelled data. The model consists of Task Specialists (T S) and Task Predictor (T P) with pre-trained Stable Diffusion (SD) module. A comparison study with four SOTA models conducted on three real-world datasets shows that the proposed model outperforms all the selected baselines.
  arXiv  Detail & Related papers  (2024-09-26T12:59:09Z)
• $k$NN Prompting: Beyond-Context Learning with Calibration-Free Nearest Neighbor Inference [75.08572535009276]
  In-Context Learning (ICL) formulates target tasks as prompt completion conditioned on in-context demonstrations. $k$NN Prompting first queries LLM with training data for distributed representations, then predicts test instances by simply referring to nearest neighbors. It significantly outperforms state-of-the-art calibration-based methods under comparable few-shot scenario.
  arXiv  Detail & Related papers  (2023-03-24T06:16:29Z)
• Pushing the Limits of Simple Pipelines for Few-Shot Learning: External Data and Fine-Tuning Make a Difference [74.80730361332711]
  Few-shot learning is an important and topical problem in computer vision. We show that a simple transformer-based pipeline yields surprisingly good performance on standard benchmarks.
  arXiv  Detail & Related papers  (2022-04-15T02:55:58Z)
• Lifelong Learning Without a Task Oracle [13.331659934508764]
  Supervised deep neural networks are known to undergo a sharp decline in the accuracy of older tasks when new tasks are learned. We propose and compare several candidate task-assigning mappers which require very little memory overhead. Best-performing variants only impose an average cost of 1.7% parameter memory increase.
  arXiv  Detail & Related papers  (2020-11-09T21:30:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.