SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.28301 · 2026

LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models

Hyunwoo Kim, Dahuin Jung, Minseok Kang, Minwoo Kim, Chanyoung Kim

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 2 no recorded run), measured 2026-11-22; 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/2603.28301.json · badge: /api/ran/2603.28301.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
cau-hai-lab/LIBERO-Para community
Papers with Code archive
4 ran · 2 no recorded run
FunctionStatusWhere it lives
build_dependency_tree Ran
ran; our contract draft was wrong, not the code
cau-hai-lab/LIBERO-Para/metrics/compute_pride_score_example.py
code served (permissive licence) · get_code("e6c29da87132a35e")
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) and returned TreeNode. 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

class Token:
    def __init__(self, dep_, pos_, children=None):
        self.dep_ = dep_
        self.pos_ = pos_
        self.children = children or []

class Doc:
    def __init__(self, tokens):
        self.tokens = tokens
    
    def __iter__(self):
        return iter(self.tokens)

# Create a simple dependency tree structure
# Token structure: dep_ (dependency type), pos_ (part of speech), children (list of child tokens)
token_leaf1 = Token(dep_='nsubj', pos_='NOUN', children=[])
token_leaf2 = Token(dep_='dobj', pos_='NOUN', children=[])
token_leaf3 = Token(dep_='prep', pos_='ADP', children=[])

token_root = Token(dep_='ROOT', pos_='VERB', children=[token_leaf1, token_leaf2, token_leaf3])

doc = Doc([token_leaf1, token_leaf2, token_leaf3, token_root])

ARGS = {
    'doc': doc
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type TreeNode — a TreeNode output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a TreeNode 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
get_leftmost_leaf_indices Ran
ran, honoured the contract we drafted
cau-hai-lab/LIBERO-Para/metrics/compute_pride_score_example.py
code served (permissive licence) · get_code("49e6f0b553894a75")
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 list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 Node:
    def __init__(self, children=None):
        self.children = children if children is not None else []

# Create a simple tree structure
# Tree:
#     0
#    / \
#   1   2
#  / \
# 3   4

node3 = Node(children=[])
node4 = Node(children=[])
node1 = Node(children=[node3, node4])
node2 = Node(children=[])
node0 = Node(children=[node1, node2])

nodes = [node0, node1, node2, node3, node4]

ARGS = {
    "nodes": nodes
}
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 list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
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
tree_to_list Ran
ran; our contract draft was wrong, not the code
cau-hai-lab/LIBERO-Para/metrics/compute_pride_score_example.py
code served (permissive licence) · get_code("dc97f5fe9c79f957")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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
from torch import nn

class TreeNode:
    def __init__(self, label, children=None):
        self.label = label
        self.children = children or []

def tree_to_list(node: TreeNode) -> list:
    result = []
    for child in node.children:
        result.extend(tree_to_list(child))
    result.append(node)
    return result

# Construct a simple tree structure
leaf1 = TreeNode(label="leaf1", children=[])
leaf2 = TreeNode(label="leaf2", children=[])
leaf3 = TreeNode(label="leaf3", children=[])
branch1 = TreeNode(label="branch1", children=[leaf1, leaf2])
root = TreeNode(label="root", children=[branch1, leaf3])

ARGS = {
    "node": root
}
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 list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
zhang_shasha_distance Ran
ran, honoured the contract we drafted
cau-hai-lab/LIBERO-Para/metrics/compute_pride_score_example.py
code served (permissive licence) · get_code("5a0ba5ac1f5d37b3")
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.

import numpy as np

class TreeNode:
    def __init__(self, label, children=None):
        self.label = label
        self.children = children or []

# Create two simple tree structures for testing
# Tree 1: root with two children
leaf1_a = TreeNode("A")
leaf1_b = TreeNode("B")
tree1 = TreeNode("root1", children=[leaf1_a, leaf1_b])

# Tree 2: root with one child that has a child
leaf2_a = TreeNode("X")
leaf2_b = TreeNode("Y")
intermediate2 = TreeNode("mid", children=[leaf2_a, leaf2_b])
tree2 = TreeNode("root2", children=[intermediate2])

ARGS = {
    "tree1": tree1,
    "tree2": tree2,
}
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 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
StructuralSimilarity Not yet run
unverified (harvested, no recorded run)
cau-hai-lab/LIBERO-Para/metrics/compute_pride_score_example.py
code served (permissive licence) · get_code("eec23c4cc6e35ae3")
How it ran

no invocation recorded.

TreeNode Not yet run
unverified (harvested, no recorded run)
cau-hai-lab/LIBERO-Para/metrics/compute_pride_score_example.py
code served (permissive licence) · get_code("ea541b08cda0e7eb")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 6 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.

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

Vision-Language-Action (VLA) models achieve strong performance in robotic manipulation by leveraging pre-trained vision-language backbones. However, in downstream robotic settings, they are typically fine-tuned with limited data, leading to overfitting to specific instruction formulations and leaving robustness to paraphrased instructions underexplored. To study this gap, we introduce LIBERO-Para, a controlled benchmark that independently varies action expressions and object references for fine-grained analysis of linguistic generalization. Across seven VLA configurations (0.6B-7.5B), we observe consistent performance degradation of 22-52 pp under paraphrasing. This degradation is primarily driven by object-level lexical variation: even simple synonym substitutions cause large drops, indicating reliance on surface-level matching rather than semantic grounding. Moreover, 80-96% of failures arise from planning-level trajectory divergence rather than execution errors, showing that paraphrasing disrupts task identification. Binary success rate treats all paraphrases equally, obscuring whether models perform consistently across difficulty levels or rely on easier cases. To address this, we propose PRIDE, a metric that quantifies paraphrase difficulty using semantic and syntactic factors. Our benchmark and corresponding code are available at: https: //github.com/cau-hai-lab/LIBERO-Para

For agents

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

get_harvested_code_for_paper("2603.28301")
get_code_for_paper("2603.28301")
have("2603.28301")

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

curl https://syntology.ai/api/ran/2603.28301.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.28301.svg)](https://syntology.ai/paper/2603.28301)

Connect an agent — have() is free.