SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.09246 · 2024

OpenVLA: An Open-Source Vision-Language-Action Model

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 3 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2406.09246.json · badge: /api/ran/2406.09246.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
openvla/openvla canonical (classifier label)
classifier basis: census
2 ran · 2 no recorded run
stanford-iliad/openvla-mini community (classifier label)
classifier basis: plan_row
4 ran · 0 no recorded run
pie33000/OpenVLA community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
crop_and_resize Ran
ran; our contract draft was wrong, not the code
openvla/openvla/experiments/robot/openvla_utils.py
code served (permissive licence) · get_code("d4883358a40b04e4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned EagerTensor, ndim 3, dtype family float, shape (224, 224, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)

# Create a small test image with shape (H, W, C) = (224, 224, 3)
# Values between [0, 1] as required
image_3d = tf.constant(np.random.rand(224, 224, 3).astype(np.float32))

# Create a batched image with shape (batch_size, H, W, C) = (2, 224, 224, 3)
image_4d = tf.constant(np.random.rand(2, 224, 224, 3).astype(np.float32))

# crop_scale: should be between 0 and 1 (represents area ratio)
crop_scale = 0.64  # This will result in sqrt(0.64) = 0.8, so 80% of original height/width

# batch_size: must match the batch dimension
batch_size = 2

# Test with 3D image (will be expanded internally)
ARGS = {
    "image": image_3d,
    "crop_scale": crop_scale,
    "batch_size": 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 3, dtype family float, shape (224, 224, 3) — shape (224, 224, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_image_resize_size Ran
ran, honoured the contract we drafted
openvla/openvla/experiments/robot/robot_utils.py
code served (permissive licence) · get_code("01fa10b1e4f2b232")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

class Config:
    def __init__(self, model_family):
        self.model_family = model_family

cfg = Config(model_family='openvla')

ARGS = {
    'cfg': cfg
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_openvla_prompt Ran
ran; our contract draft was wrong, not the code
stanford-iliad/openvla-mini/vla-scripts/extern/verify_openvla.py
code served (permissive licence) · get_code("883da746d4398ec2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    'instruction': 'pick up the cube'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
hr_name Ran
ran on a synthesized input
stanford-iliad/openvla-mini/experiments/robot/robot_utils.py
code served (permissive licence) · get_code("ef2c04230ed3e692")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math

def round_to_n(x, n=1):
    return round(x, -int(math.floor(math.log10(abs(x))) + (n - 1)))

def hr_name(float_arg, fp=None):
    if fp is not None:
        float_arg = round_to_n(float_arg, n=fp)
    return str(float_arg).replace('.', '_')

ARGS = {
    'float_arg': 0.0001234,
    'fp': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_stanford_hydra_dataset Ran
ran on a synthesized input
pie33000/OpenVLA/convert_tf_to_numpy.py
code served (permissive licence) · get_code("81047a80f05792ed")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Construct a valid step dict that matches what parse_stanford_hydra_dataset expects
step = {
    'observation': {
        'image': torch.randn(2, 3, 4, 4)  # batch=2, channels=3, height=4, width=4
    },
    'action': torch.randn(2, 5),  # batch=2, action_dim=5
    'language_instruction': b'pick up the red cube'  # bytes that can be decoded
}

ARGS = {
    'step': step
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_viola_dataset Ran
ran on a synthesized input
pie33000/OpenVLA/convert_tf_to_numpy.py
code served (permissive licence) · get_code("6595ff27135d42d4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct the step dict that parse_viola_dataset expects
step = {
    'observation': {
        'agentview_rgb': np.random.randint(0, 256, (64, 64, 3), dtype=np.uint8),
        'natural_language_instruction': b'pick up the object'
    },
    'action': {
        'world_vector': np.array([0.1, 0.2, 0.3]),
        'rotation_delta': np.array([0.01, 0.02, 0.03, 0.04]),
        'gripper_closedness_action': 0.5
    }
}

ARGS = {
    'step': step
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
remap_state_dicts_for_hf Ran
ran; our contract draft was wrong, not the code
stanford-iliad/openvla-mini/vla-scripts/extern/convert_openvla_weights_to_hf.py
code served (permissive licence) · get_code("7af07bf1a8a753fe")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create tiny state dicts with the expected key structures
prismatic_vision_backbone_state_dict = {
    'featurizer.layer1.weight': torch.randn(4, 3),
    'featurizer.layer1.bias': torch.randn(4),
    'featurizer.layer2.weight': torch.randn(5, 4),
}

projector_state_dict = {
    'projector.0.weight': torch.randn(6, 5),
    'projector.0.bias': torch.randn(6),
    'projector.2.weight': torch.randn(7, 6),
    'projector.2.bias': torch.randn(7),
    'projector.4.weight': torch.randn(8, 7),
    'projector.4.bias': torch.randn(8),
}

llm_backbone_state_dict = {
    'llm.layer1.weight': torch.randn(9, 8),
    'llm.layer1.bias': torch.randn(9),
    'llm.layer2.weight': torch.randn(10, 9),
}

ARGS = {
    'prismatic_vision_backbone_state_dict': prismatic_vision_backbone_state_dict,
    'projector_state_dict': projector_state_dict,
    'llm_backbone_state_dict': llm_backbone_state_dict,
    'use_fused_vision_backbone': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
round_to_n Ran
ran on a synthesized input
stanford-iliad/openvla-mini/experiments/robot/robot_utils.py
code served (permissive licence) · get_code("a6a8431fbd02a85e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math

def round_to_n(x, n=1):
    return round(x, -int(math.floor(math.log10(abs(x))) + (n - 1)))

ARGS = {
    "x": 12345.6789,
    "n": 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_openvla_prompt Not yet run
unverified (harvested, no recorded run)
openvla/openvla/vla-scripts/deploy.py
code served (permissive licence) · get_code("5dcd468341e1592d")
How it ran

no invocation recorded.

get_processor Not yet run
unverified (harvested, no recorded run)
openvla/openvla/experiments/robot/openvla_utils.py
code served (permissive licence) · get_code("886ca388dfb17143")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 6 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 later) and 4 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
crop_and_resize 2 other papers
1 later, 1 undated
2505.06111 (2025-05)
UniVLA: Learning to Act Anywhere with Task-centric Latent Actions
openreview_7MlfE2Da2W
SCALE: Self-uncertainty Conditioned Adaptive Looking and Execution for Vision-Language-Action Models
remap_state_dicts_for_hf 2 other papers
2 later
2505.06111 (2025-05)
UniVLA: Learning to Act Anywhere with Task-centric Latent Actions
2603.04639 (2026-03)
RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies
get_image_resize_size 1 other paper
1 undated
no dated paper among those named
openreview_7MlfE2Da2W
SCALE: Self-uncertainty Conditioned Adaptive Looking and Execution for Vision-Language-Action Models
get_openvla_prompt 1 other paper
1 later
2603.04639 (2026-03)
RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies
get_openvla_prompt 1 other paper
1 later
2503.24278 (2025-03)
AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real World
get_processor 1 other paper
1 undated
no dated paper among those named
openreview_7MlfE2Da2W
SCALE: Self-uncertainty Conditioned Adaptive Looking and Execution for Vision-Language-Action Models

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Large policies pretrained on a combination of Internet-scale vision-language data and diverse robot demonstrations have the potential to change how we teach robots new skills: rather than training new behaviors from scratch, we can fine-tune such vision-language-action (VLA) models to obtain robust, generalizable policies for visuomotor control. Yet, widespread adoption of VLAs for robotics has been challenging as 1) existing VLAs are largely closed and inaccessible to the public, and 2) prior work fails to explore methods for efficiently fine-tuning VLAs for new tasks, a key component for adoption. Addressing these challenges, we introduce OpenVLA, a 7B-parameter open-source VLA trained on a diverse collection of 970k real-world robot demonstrations. OpenVLA builds on a Llama 2 language model combined with a visual encoder that fuses pretrained features from DINOv2 and SigLIP. As a product of the added data diversity and new model components, OpenVLA demonstrates strong results for generalist manipulation, outperforming closed models such as RT-2-X (55B) by 16.5% in absolute task success rate across 29 tasks and multiple robot embodiments, with 7x fewer parameters. We further show that we can effectively fine-tune OpenVLA for new settings, with especially strong generalization results in multi-task environments involving multiple objects and strong language grounding abilities, and outperform expressive from-scratch imitation learning methods such as Diffusion Policy by 20.4%. We also explore compute efficiency; as a separate contribution, we show that OpenVLA can be fine-tuned on consumer GPUs via modern low-rank adaptation methods and served efficiently via quantization without a hit to downstream success rate. Finally, we release model checkpoints, fine-tuning notebooks, and our PyTorch codebase with built-in support for training VLAs at scale on Open X-Embodiment datasets.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2406.09246")
get_code_for_paper("2406.09246")
have("2406.09246")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2406.09246.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2406.09246.svg)](https://syntology.ai/paper/2406.09246)

Connect an agent — have() is free.