# Show HN: VLDA-learn compact Deltas as actionable transition interfaces for robot > Source: > Published: 2026-10-22 12:30:11+00:00 # LoongForge-VLDA **Vision-Language-Delta-Action:** learning compact Deltas as actionable transition interfaces for robotic manipulation. 1 Harbin Institute of Technology, Shenzhen  ·  2 Baidu, Inc.  ·  3 Shenzhen Loop Area Institute * Equal contribution † Corresponding author **98.5%** LIBERO average **84.5%** LIBERO-Plus, zero-shot **62.9%** RoboTwin 2.0, 10 hard tasks **73.0%** Real robot · R1 Lite **72.0%** Real robot · AgileX **1 GPU** Real-world inference on a single GPU **01** The idea ## Don't predict the whole future. *Capture the change.* **What does an action actually change?** Manipulation is a chain of state transitions, and most of every frame stays the same. VLA models must infer that change implicitly; world-action models predict whole future frames and spend capacity on pixels that never move. VLDA distills only **what changes** into a compact Delta and hands it to the action head as an explicit interface: understand the scene, express the transition, then execute. **Understand** Vision and language are encoded into a task-aware representation. **Transition · Δ** A compact Delta captures the local and global change the task requires. **Execute** The action head is refined by the Delta, with no future frames needed at inference. **02** The method ## From a Delta to an action, *in three moves.* A V-JEPA 2 EMA teacher provides the target transition. Two filters distill it into a Delta; one refinement module feeds it back into the action head. ### Local Transition Filter Selects the top-64 regions with the largest transition magnitude and matches them to the predicted Delta by Sinkhorn optimal transport, so the Delta is not diluted by static background. ### Global Transition Anchor Anchors the local patches to a global summary of the transition, keeping the Delta consistent with the task as a whole. ### Delta-Driven Action Refinement **DAMG** grounds the Delta on the action manifold with an InfoNCE loss. **DSAF** seeds the action flow with a Delta-based noise bias and a prefix token. **Learn the transition** Pre-train the Delta from the teacher, without actions. `L1 = L_LTF + L_GTA` **Learn to act** Train the action head with the Delta as an interface, keeping the stage-1 loss as a regularizer. `L2 = L_act + λ·L1 + λ_DDAR·L_DDAR` **Built on LoongForge.** VLDA is trained with the LoongForge embodied stack: DDP, ZeRO-1, FSDP and HSDP, component-wise freezing and checkpointing. Training used 8 GPUs. Inference needs no future observations, and the EMA teacher can be dropped after training. **03** The results ## Stronger where it counts, *consistently.* Three simulation benchmarks and a set of ablations. Our row is highlighted. **98.5%** LIBERO average ### Top average, with the biggest lift on long-horizon tasks. VLDA beats π0.5 by 1.6 points on average and edges out every other baseline, reaching 95.2% on the Long suite. Success rate (%) on LIBERO | Method | Spatial | Object | Goal | Long | Avg | |---|---|---|---|---|---| | VLDA | 99.8 | 99.6 | 99.4 | 95.2 | 98.5 | | π0.5 | 98.8 | 98.2 | 98.0 | 92.4 | 96.9 | | VLA-Adapter | 97.8 | 99.2 | 97.2 | 95.0 | 97.3 | | OpenVLA-OFT | 97.6 | 98.4 | 97.9 | 94.5 | 97.1 | | CronusVLA | 97.3 | 99.6 | 96.9 | 94.0 | 97.0 | | InternVLA-M1 | 98.0 | 99.0 | 93.8 | 92.6 | 95.9 | | UniVLA | 95.4 | 98.8 | 93.6 | 94.0 | 95.4 | | π0 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | | GR00T-N1 | 94.4 | 97.6 | 93.0 | 90.6 | 93.9 | **+10.9** points over π0.5, zero-shot ### Robust when the world is perturbed. Ahead of π0.5 on all seven perturbation dimensions, and ahead of the strongest baseline, GuidedVLA, on six of them (+9.1 on average). The gap widens on the Long suite: +13.4 over π0.5 and +12.1 over GuidedVLA. Zero-shot success rate (%) on LIBERO-Plus | Method | Camera | Robot | Language | Light | Background | Noise | Layout | Avg | |---|---|---|---|---|---|---|---|---| | VLDA | 72.4 | 75.3 | 84.0 | 96.0 | 95.5 | 89.4 | 85.7 | 84.5 | | GuidedVLA | 73.7 | 51.4 | 62.6 | 94.6 | 89.0 | 85.2 | 79.9 | 75.4 | | π0.5 | 59.7 | 65.5 | 75.3 | 87.0 | 82.4 | 72.1 | 80.3 | 73.6 | | DreamVLA | 65.0 | 40.8 | 63.5 | 85.7 | 82.6 | 84.9 | 74.0 | 69.9 | | OpenVLA-OFT | 56.4 | 31.9 | 79.5 | 88.7 | 93.3 | 75.8 | 74.2 | 69.6 | | π0 | 13.8 | 6.0 | 58.8 | 85.0 | 81.4 | 79.0 | 68.9 | 53.6 | **+20.2** points over π0.5, 10 hard tasks ### Biggest gains on complex state-transition tasks. Best or tied-best on every selected task. Put Bottles Dustbin jumps +46, Stacking Three +30, Place Dual Shoes +27 and Open Microwave +27. Across all 50 tasks VLDA scores 50.86% against 43.84% for π0.5. Success rate (%) on RoboTwin 2.0 | Method | Ranking RGB | Ranking Size | Grab Roller | Open Laptop | Open Microwave | Place Dual Shoes | Put Bottles Dustbin | Shake Bottle | Stacking Three | Stacking Two | Avg | |---|---|---|---|---|---|---|---|---|---|---|---| | VLDA | 53 | 16 | 96 | 60 | 64 | 34 | 55 | 100 | 65 | 86 | 62.9 | | π0.5 | 35 | 14 | 89 | 35 | 37 | 7 | 9 | 100 | 35 | 66 | 42.7 | | π0 | 5 | 1 | 80 | 46 | 50 | 0 | 13 | 60 | 24 | 41 | 32.0 | | RDT | 0 | 0 | 43 | 32 | 20 | 4 | 4 | 45 | 17 | 30 | 19.5 | **+24** points on Basket Relocation ### Every module earns its place. Removing any single piece costs accuracy. LT-Filter and GT-Anchor each help alone, and together they help most, notably on the real robot. Ablations (LIBERO average unless noted) ~~98.5~~ **97.2** ~~98.5~~ **97.3** ~~98.5~~ **97.8** ~~84.5~~ **82.0** ### LT-Filter and GT-Anchor | LT-Filter | GT-Anchor | LIBERO Spatial | Object | Goal | Long | Ring | Sweep | Sort | Basket | |---|---|---|---|---|---|---|---|---|---| | ✗ | ✗ | 98.6 | 98.4 | 98.0 | 92.8 | 76 | 52 | 48 | 48 | | ✓ | ✗ | 99.4 | 99.4 | 99.0 | 94.4 | 84 | 60 | 56 | 68 | | ✗ | ✓ | 98.8 | 99.0 | 98.8 | 93.8 | 80 | 60 | 52 | 60 | | ✓ | ✓ | 99.8 | 99.6 | 99.4 | 95.2 | 92 | 68 | 60 | 72 | Simulation columns are LIBERO success rate (%); real-world columns are Ring Stacking, Table Sweeping, Conveyor Sorting and Basket Relocation on R1 Lite. Each module helps alone; together they help most. **04** In the real world ## Still ahead on *real robots.* Four long-horizon tasks on a GALAXEA R1 Lite, 25 trials each, plus a second platform, AgileX Cobot Magic. Numbers are final-stage success rates (%). #### Ring Stacking **92%** vs. π0.5 80% #### Table Sweeping **68%** vs. π0.5 52% #### Conveyor Sorting **60%** vs. π0.5 44% #### Basket Relocation **72%** vs. π0.5 48% **+17.0** points over π0.5, R1 Lite Averaged over four tasks, VLDA reaches 73.0% against 56.0% for π0.5. On a second robot the lead holds. Basket Relocation uses an out-of-distribution mat position, and VLDA still reaches 72% at the final placement. | Platform (average) | OpenVLA-OFT | π0.5 | VLDA | |---|---|---|---| | GALAXEA R1 Lite | 16.0 | 56.0 | 73.0 | | AgileX Cobot Magic | 17.0 | 54.0 | 72.0 | | Δ vs. π0.5 (pp) | | | +17.0 / +18.0 | ### What the Delta looks at The Delta concentrates on the regions that move and interact, not on the static scene. **05** Build with VLDA ## Code and *citation.* [GitHub ↗](https://github.com/baidu-baige/LoongForge/tree/master/cookbook/loongforge-vlda) ### LoongForge-VLDA Cookbook Training code, configs and recipes for VLDA live in LoongForge. ``` git clone https://github.com/baidu-baige/LoongForge.git ``` ### Citation ``` @misc{zhang2026visionlanguagedeltaactionlearningcompactdeltas, title={Vision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic Manipulation}, author={Lingsen Zhang and Wei Li and Jie He and Henghua Zhang and Kai Mo and Shiyong Li and Yanpeng Wang and Rui Shao}, year={2026}, eprint={2610.25874}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2610.25874}, } ``` Read the preprint on [arXiv ↗](https://arxiv.org/abs/2610.25874).