SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.03967 · EMNLP · 2024

Investigating the Role of Instruction Variety and Task Difficulty in Robotic Manipulation Tasks

Alessandro Suglia, Ioannis Konstas, Nikolas Vitsakis, Amit Parekh

arXiv · PDF · Open in the Atlas

Code that ran

8 of 16 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 failures of our instrument; 8 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/2407.03967.json · badge: /api/ran/2407.03967.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
amitkparekh/CoGeLoT canonical (classifier label)
classifier basis: census
8 ran · 8 no recorded run
FunctionStatusWhere it lives
convert_to_dotlist Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/config.py
code served (permissive licence) · get_code("37474637f4426121")
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
import torch.nn as nn
from typing import Any

# Construct a valid config dictionary with various types of values
config = {
    "learning_rate": 0.001,
    "batch_size": 32,
    "model_name": "bert-base",
    "dropout": 0.1,
    "use_cache": True,
    "layers": 12,
    "hidden_size": 768,
    "max_seq_length": None,  # This will be converted to 'null'
    "optimizer": "adam",
    "epochs": 10,
    "checkpoint_path": None,  # This will also be converted to 'null'
}

ARGS = {
    "config": config
}
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
flatten_config Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/config.py
code served (permissive licence) · get_code("95db24d20b373cb0")
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.

from omegaconf import DictConfig, OmegaConf

# Create a sample nested DictConfig for testing
config_dict = {
    "model": {
        "name": "resnet50",
        "num_layers": 50,
        "pretrained": True
    },
    "training": {
        "learning_rate": 0.001,
        "batch_size": 32,
        "epochs": 100
    },
    "data": {
        "dataset": "imagenet",
        "split": "train"
    }
}

config = OmegaConf.create(config_dict)

ARGS = {
    "config": config,
    "separator": "."
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
omegaconf, pandas — 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
load_json Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/io.py
code served (permissive licence) · get_code("cc6276625cb68fc3")
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 tempfile
import gzip
from pathlib import Path
import orjson

# Create a temporary directory for test files
temp_dir = tempfile.mkdtemp()

# Create test data
test_data = {"key": "value", "number": 42, "list": [1, 2, 3]}

# Write uncompressed JSON file
json_path_uncompressed = Path(temp_dir) / "test.json"
json_path_uncompressed.write_bytes(orjson.dumps(test_data))

# Write compressed JSON file (.json.gz)
json_path_compressed = Path(temp_dir) / "test.json.gz"
compressed_bytes = gzip.compress(orjson.dumps(test_data))
json_path_compressed.write_bytes(compressed_bytes)

# Test with uncompressed file
ARGS = {
    "path": json_path_uncompressed
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
orjson — 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
maybe_split_dataset_by_node Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/hf_datasets.py
code served (permissive licence) · get_code("2532183c09141f04")
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 Dataset. 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 os
import datasets
import numpy as np

# Create a simple dataset that works with the function
data = {
    'text': ['hello world', 'foo bar', 'baz qux'],
    'label': [0, 1, 0]
}
dataset = datasets.Dataset.from_dict(data)

# Construct arguments
ARGS = {
    'dataset': dataset
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
datasets, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Dataset — a Dataset output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Dataset output the recorded ndim 1 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
orjson_dumps Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/io.py
code served (permissive licence) · get_code("3d5937c99effbab1")
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
import numpy as np
import orjson
from typing import Any

torch.manual_seed(0)

# Create a default function that handles non-serializable types
def default_handler(obj: Any) -> Any:
    """Handle non-serializable types for orjson."""
    if isinstance(obj, torch.Tensor):
        return obj.cpu().numpy().tolist()
    elif isinstance(obj, np.ndarray):
        return obj.tolist()
    elif hasattr(obj, '__dict__'):
        return obj.__dict__
    else:
        return str(obj)

# Create test data with various types that need serialization
test_data = {
    "tensor": torch.tensor([1.0, 2.0, 3.0, 4.0]),
    "numpy_array": np.array([1, 2, 3, 4]),
    "list": [1, 2, 3],
    "dict": {"nested": "value"},
    "int": 42,
    "float": 3.14,
    "string": "hello",
    123: "numeric_key"  # non-string key - requires OPT_NON_STR_KEYS
}

ARGS = {
    "v": test_data,
    "default": default_handler
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, orjson, 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
remove_hydra_key_from_config Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/hydra.py
code served (permissive licence) · get_code("23746069d63af15f")
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 DictConfig. 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.

from omegaconf import DictConfig, OmegaConf

# Create a minimal DictConfig with a hydra key
config_dict = {
    'hydra': {
        'run': {'dir': 'outputs'},
        'sweep': {'dir': 'multirun'}
    },
    'model': {
        'name': 'test_model',
        'hidden_size': 128
    },
    'data': {
        'batch_size': 32,
        'num_workers': 4
    }
}

config = OmegaConf.create(config_dict)

ARGS = {
    'config': config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
omegaconf — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DictConfig — a DictConfig output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DictConfig output the recorded ndim 1 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
rewire_trainer_callbacks Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/hydra.py
code served (permissive licence) · get_code("1c3c7294b2a29498")
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 DictConfig. 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.

from omegaconf import DictConfig, OmegaConf

# Create a config with trainer.callbacks as a mapping (dict)
config = OmegaConf.create({
    "trainer": {
        "callbacks": {
            "callback_1": {"name": "checkpoint"},
            "callback_2": {"name": "early_stopping"},
            "callback_3": {"name": "lr_monitor"}
        }
    }
})

ARGS = {
    "config": config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
omegaconf — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DictConfig — a DictConfig output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DictConfig output the recorded ndim 1 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
save_json Ran
ran; our contract draft was wrong, not the code
amitkparekh/CoGeLoT/src/cogelot/common/io.py
code served (permissive licence) · get_code("539e6765949c8216")
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 PosixPath. 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 tempfile
from pathlib import Path
import numpy as np

# Create a temporary directory for the test
temp_dir = tempfile.mkdtemp()
temp_path = Path(temp_dir) / "test_data.json"

# Prepare test data - use a mix of types that orjson can serialize
test_data = {
    "strings": ["hello", "world"],
    "numbers": [1, 2, 3, 4.5],
    "nested": {"key": "value"},
    "numpy_array": np.array([1, 2, 3, 4]),
    "numpy_float": np.float32(3.14),
}

ARGS = {
    "data": test_data,
    "path": temp_path,
    "compress": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, orjson — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
build_eval_run_name Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/config_metadata_patcher.py
code served (permissive licence) · get_code("6001ce21d2b2a41c")
How it ran

no invocation recorded.

download_model_checkpoint Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/hf_models.py
code served (permissive licence) · get_code("b0f34e18863804c5")
How it ran

no invocation recorded.

download_model_from_wandb Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/wandb.py
code served (permissive licence) · get_code("eb78040bd689b352")
How it ran

no invocation recorded.

get_location_of_hub_parquet_files Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/hf_datasets.py
code served (permissive licence) · get_code("bbd4f08e451b847d")
How it ran

no invocation recorded.

get_model_checkpoint_file_in_remote_repo_for_epoch Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/hf_models.py
code served (permissive licence) · get_code("e9ec2c6761040369")
How it ran

no invocation recorded.

load_dataset_from_files Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/hf_datasets.py
code served (permissive licence) · get_code("b3415cf1d15a79d3")
How it ran

no invocation recorded.

update_eval_config Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/config_metadata_patcher.py
code served (permissive licence) · get_code("07b532469f51b9d4")
How it ran

no invocation recorded.

upload_model_checkpoint Not yet run
unverified (harvested, no recorded run)
amitkparekh/CoGeLoT/src/cogelot/common/hf_models.py
code served (permissive licence) · get_code("bbb9da8a2d088a4c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 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 16 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

Evaluating the generalisation capabilities of multimodal models based solely on their performance on out-of-distribution data fails to capture their true robustness. This work introduces a comprehensive evaluation framework that systematically examines the role of instructions and inputs in the generalisation abilities of such models, considering architectural design, input perturbations across language and vision modalities, and increased task complexity. The proposed framework uncovers the resilience of multimodal models to extreme instruction perturbations and their vulnerability to observational changes, raising concerns about overfitting to spurious correlations. By employing this evaluation framework on current Transformerbased multimodal models for robotic manipulation tasks, we uncover limitations and suggest future advancements should focus on architectural and training innovations that better integrate multimodal inputs, enhancing a model's generalisation prowess by prioritising sensitivity to input content over incidental correlations. 1 L1 L2 L3 L4 (a) Trained on Original; Evaluated on Original

For agents

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

get_harvested_code_for_paper("2407.03967")
get_code_for_paper("2407.03967")
have("2407.03967")

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

curl https://syntology.ai/api/ran/2407.03967.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.03967.svg)](https://syntology.ai/paper/2407.03967)

Connect an agent — have() is free.