CrossVLA: Cross-Paradigm Post-Training and Inference Optimization for Vision-Language-Action Models
Vision-Language-Action (VLA) models have rapidly converged on a small set of architectural patterns: discrete-token autoregression (e.g. OpenVLA) and continuous-action flow-matching (e.g. pi-0.5). Yet preference alignment via Direct Preference Optimisation (DPO) -- the de-facto post-training step in language models -- has been studied almost exclusively on autoregressive VLAs. We present CrossVLA, an empirical study of cross-paradigm VLA post-training. Three contributions: (i) a surrogate flow-matching log-probability estimator that lets DPO operate on continuous-action backbones without probability-flow ODE integration; (ii) a head-to-head comparison of LoRA and DoRA as the parameter-efficient layer for VLA DPO, finding DoRA improves over OpenVLA SFT by a mean +10.4 pp across LIBERO 4-suite (600 trials, 3 seeds) -- per-suite +20.0 Object, +11.0 Long-horizon, +8.0 Goal, +2.7 Spatial -- with zero seed variance on Object (38/50 on each of 3 seeds); (iii) an inference-time anatomy showing the denoise loop dominates 78.6% of sample_actions latency and prefix-K/V caching a la VLA-Cache caps at a 21% acceleration ceiling -- both chunk-level and token-level cache strategies degrade success rate to 0-80% in our benchmarks. We further pretrain a multi-view + temporal projection head on 6000 LIBERO frames, achieving 99.5% k-NN recall@1 for same-task retrieval (36x over random), available as a downstream initialisation. All code, ckpts, training logs, and reproduction scripts are open at https://github.com/lz-googlefycy/vla-lab.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
One Object, Multiple Lies: A Benchmark for Cross-task Adversarial Attack on Unified Vision-Language Models
Unified vision-language models(VLMs) have recently shown remarkable progress, enabling a single model to flexibly address diverse tasks through different instructions within a shared computational architecture. This inst…
Adversarial AttackSailing AI by the Stars: A Survey of Learning from Rewards in Post-Training and Test-Time Scaling of Large Language Models
Recent developments in Large Language Models (LLMs) have shifted from pre-training scaling to post-training and test-time scaling. Across these developments, a key unified paradigm has arisen: Learning from Rewards, wher…
Active LearningMindGPT-4ov: An Enhanced MLLM via a Multi-Stage Post-Training Paradigm
We present MindGPT-4ov, a multimodal large language model (MLLM) that introduces a general post-training paradigm spanning data production, model training, and efficient deployment. It achieves state-of-the-art performan…
Reinforcement LearningDomain AdaptationPsi-Sampler: Initial Particle Sampling for SMC-Based Inference-Time Reward Alignment in Score Models
We introduce $\Psi$-Sampler, an SMC-based framework incorporating pCNL-based initial particle sampling for effective inference-time reward alignment with a score-based generative model. Inference-time reward alignment wi…
DenoisingImage GenerationLayout-to-Image GenerationASPIRE: Iterative Amortized Posterior Inference for Bayesian Inverse Problems
Due to their uncertainty quantification, Bayesian solutions to inverse problems are the framework of choice in applications that are risk averse. These benefits come at the cost of computations that are in general, intra…
Uncertainty QuantificationVariational Inference