SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.13009 · AAAI · 2021

UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark

Ronan Le Bras, Chandra Bhagavatula, Yejin Choi, Nicholas Lourie

arXiv · PDF · Open in the Atlas

Code that ran

3 of 5 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 3 have no contract verdict; 0 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/2103.13009.json · badge: /api/ran/2103.13009.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
allenai/rainbow canonical (classifier label)
classifier basis: census
3 ran · 2 no recorded run
FunctionStatusWhere it lives
make_add_field_names_preprocessor Ran
ran on a synthesized input
allenai/rainbow/src/rainbow/preprocessors.py
code served (permissive licence) · get_code("21730ebd53e72357")
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 function. 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 tensorflow as tf
import numpy as np

# Create sample data - a dataset of lists/tuples of tensors
def create_sample_dataset():
    # Create a simple dataset with 2 samples, each having 3 tensors
    data = [
        (tf.constant([1.0, 2.0, 3.0, 4.0]), tf.constant([5.0, 6.0, 7.0, 8.0]), tf.constant([9.0, 10.0, 11.0, 12.0])),
        (tf.constant([13.0, 14.0, 15.0, 16.0]), tf.constant([17.0, 18.0, 19.0, 20.0]), tf.constant([21.0, 22.0, 23.0, 24.0]))
    ]
    dataset = tf.data.Dataset.from_generator(
        lambda: iter(data),
        output_signature=(
            tf.TensorSpec(shape=(4,), dtype=tf.float32),
            tf.TensorSpec(shape=(4,), dtype=tf.float32),
            tf.TensorSpec(shape=(4,), dtype=tf.float32)
        )
    )
    return dataset

# Test case 1: With field_names only (field_indices=None)
field_names_1 = ["input", "target", "weight"]

# Test case 2: With field_names and explicit field_indices
field_names_2 = ["a", "b", "c"]
field_indices_2 = [0, 1, 2]

ARGS = {
    "field_names": field_names_1,
    "field_indices": None
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
make_filter_preprocessor Ran
ran on a synthesized input
allenai/rainbow/src/rainbow/preprocessors.py
code served (permissive licence) · get_code("0d51b04143d199db")
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 function. 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 tensorflow as tf

# Define a simple predicate function that filters examples
# The predicate takes an example and returns a boolean
def simple_predicate(example):
    """Example predicate that keeps examples where value > 5"""
    # Assuming example is a tensor or dict with numeric data
    if isinstance(example, dict):
        return example.get('value', 0) > 5
    else:
        return example > 5

ARGS = {
    'predicate': simple_predicate
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
string_to_seed Ran
ran on a synthesized input
allenai/rainbow/src/rainbow/utils.py
code served (permissive licence) · get_code("bba4a76bdd42388e")
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 hashlib

def string_to_seed(s: str) -> int:
    """Return an integer suitable for using as a random seed.

    Given a string, ``s``, return an integer suitable for using as a
    random seed in a deterministic way based on ``s``.

    Parameters
    ----------
    s : str, required
        The string to convert into an integer.
    """
    checksum = hashlib.sha256(s.encode('utf-8')).hexdigest()
    seed = int(checksum[:8], 16)
    return seed


ARGS = {
    "s": "test_string_for_seed"
}
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
configure_logging Not yet run
unverified (harvested, no recorded run)
allenai/rainbow/src/rainbow/utils.py
code served (permissive licence) · get_code("40c05820ad047307")
How it ran

no invocation recorded.

data Not yet run
unverified (harvested, no recorded run)
allenai/rainbow/src/rainbow/utils.py
code served (permissive licence) · get_code("86dcf04053de697c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 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 5 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

Commonsense AI has long been seen as a near impossible goal-until recently. Now, research interest has sharply increased with an influx of new benchmarks and models. We propose two new ways to evaluate commonsense models, emphasizing their generality on new tasks and building on diverse, recently introduced benchmarks. First, we propose a new multitask benchmark, RAINBOW, to promote research on commonsense models that generalize well over multiple tasks and datasets. Second, we propose a novel evaluation, the cost equivalent curve, that sheds new insight on how the choice of source datasets, pretrained language models, and transfer learning methods impacts performance and data efficiency. We perform extensive experiments-over 200 experiments encompassing 4800 models-and report multiple valuable and sometimes surprising findings, e.g., that transfer almost always leads to better or equivalent performance if following a particular recipe, that QA-based commonsense datasets transfer well with each other, while commonsense knowledge graphs do not, and that perhaps counter-intuitively, larger models benefit more from transfer than smaller ones. Last but not least, we introduce a new universal commonsense reasoning model, UNICORN, that establishes new state-of-the-art performance across 8 popular commonsense benchmarks, αNLI (→87.3%), COSMOSQA (→91.8%), HELLASWAG (→93.9%), PIQA (→90.1%), SOCIALIQA (→83.2%), WINOGRANDE (→86.6%), CYCIC (→94.0%) and COMMONSENSEQA (→79.3%).

For agents

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

get_harvested_code_for_paper("2103.13009")
get_code_for_paper("2103.13009")
have("2103.13009")

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

curl https://syntology.ai/api/ran/2103.13009.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.13009.svg)](https://syntology.ai/paper/2103.13009)

Connect an agent — have() is free.