SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1911.03429 · 2019

ERASER: A Benchmark to Evaluate Rationalized NLP Models

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 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; 3 failures of our instrument; 0 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/1911.03429.json · badge: /api/ran/1911.03429.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
jayded/eraserbenchmark canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
annotations_from_jsonl Ran
ran; our contract draft was wrong, not the code
jayded/eraserbenchmark/rationale_benchmark/utils.py
code served (permissive licence) · get_code("2843042d2c0e6f50")
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 json
import tempfile
import os
from dataclasses import dataclass
from typing import Set, Tuple, Union, FrozenSet

@dataclass(eq=True, frozen=True)
class Evidence:
    text: Union[str, Tuple[int], Tuple[str]]
    docid: str
    start_token: int = -1
    end_token: int = -1
    start_sentence: int = -1
    end_sentence: int = -1

@dataclass(eq=True, frozen=True)
class Annotation:
    annotation_id: str
    query: Union[str, Tuple[int]]
    evidences: Union[Set[Tuple[Evidence]], FrozenSet[Tuple[Evidence]]]
    classification: str
    query_type: str = None
    docids: Set[str] = None

# Create a temporary JSONL file with valid data
temp_fd, temp_path = tempfile.mkstemp(suffix='.jsonl')
os.close(temp_fd)

# Build sample evidence groups
ev1 = Evidence(
    text="The quick brown fox",
    docid="doc1",
    start_token=0,
    end_token=4,
    start_sentence=0,
    end_sentence=1
)

ev2 = Evidence(
    text="jumps over the lazy dog",
    docid="doc2",
    start_token=0,
    end_token=5,
    start_sentence=0,
    end_sentence=1
)

# Create sample annotation records in JSONL format
annotations_data = [
    {
        "annotation_id": "ann_001",
        "query": "What does the fox do?",
        "evidences": [
            [
                {
                    "text": "The quick brown fox",
                    "docid": "doc1",
                    "start_token": 0,
                    "end_token": 4,
                    "start_sentence": 0,
                    "end_sentence": 1
                }
            ]
        ],
        "classification": "SUPPORTS",
        "query_type": "factual",
        "docids": ["doc1"]
    },
    {
        "annotation_id": "ann_002",
        "query": "What is the animal?",
        "evidences": [
            [
                {
                    "text": "jumps over the lazy dog",
                    "docid": "doc2",
                    "start_token": 0,
                    "end_token": 5,
                    "start_sentence": 0,
                    "end_sentence": 1
                }
            ]
        ],
        "classification": "REFUTES",
        "query_type": "factual",
        "docids": ["doc2"]
    }
]

# Write to temp file
with open(temp_path, 'w') as f:
    for record in annotations_data:
        f.write(json.dumps(record) + '\n')

ARGS = {
    'fp': temp_path
}
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 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
load_datasets Ran
ran; our contract draft was wrong, not the code
jayded/eraserbenchmark/rationale_benchmark/utils.py
code served (permissive licence) · get_code("e1878c4a3bf022ae")
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 json
import os
import tempfile
from dataclasses import dataclass
from typing import Set, Tuple, Union, FrozenSet

@dataclass(eq=True, frozen=True)
class Evidence:
    text: Union[str, Tuple[int], Tuple[str]]
    docid: str
    start_token: int = -1
    end_token: int = -1
    start_sentence: int = -1
    end_sentence: int = -1

@dataclass(eq=True, frozen=True)
class Annotation:
    annotation_id: str
    query: Union[str, Tuple[int]]
    evidences: Union[Set[Tuple[Evidence]], FrozenSet[Tuple[Evidence]]]
    classification: str
    query_type: str = None
    docids: Set[str] = None

# Create a temporary directory with train.jsonl, val.jsonl, test.jsonl
temp_dir = tempfile.mkdtemp()

# Create sample Evidence objects
evidence1 = Evidence(
    text="This is evidence text",
    docid="doc1",
    start_token=0,
    end_token=5,
    start_sentence=0,
    end_sentence=1
)

evidence2 = Evidence(
    text="Another evidence",
    docid="doc2",
    start_token=1,
    end_token=3,
    start_sentence=0,
    end_sentence=1
)

# Create sample Annotation objects
annotation1 = {
    "annotation_id": "ann1",
    "query": "What is this?",
    "evidences": [[
        {
            "text": "Evidence text 1",
            "docid": "doc1",
            "start_token": 0,
            "end_token": 5,
            "start_sentence": 0,
            "end_sentence": 1
        }
    ]],
    "classification": "SUPPORTS",
    "query_type": "factoid",
    "docids": ["doc1"]
}

annotation2 = {
    "annotation_id": "ann2",
    "query": "Why?",
    "evidences": [[
        {
            "text": "Evidence text 2",
            "docid": "doc2",
            "start_token": 1,
            "end_token": 3,
            "start_sentence": 0,
            "end_sentence": 1
        }
    ]],
    "classification": "REFUTES",
    "query_type": "factoid",
    "docids": ["doc2"]
}

# Write train.jsonl
with open(os.path.join(temp_dir, 'train.jsonl'), 'w') as f:
    f.write(json.dumps(annotation1) + '\n')
    f.write(json.dumps(annotation2) + '\n')

# Write val.jsonl
with open(os.path.join(temp_dir, 'val.jsonl'), 'w') as f:
    f.write(json.dumps(annotation1) + '\n')

# Write test.jsonl
with open(os.path.join(temp_dir, 'test.jsonl'), 'w') as f:
    f.write(json.dumps(annotation2) + '\n')

ARGS = {
    'data_dir': temp_dir
}
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 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
load_jsonl Ran
ran; our contract draft was wrong, not the code
jayded/eraserbenchmark/rationale_benchmark/utils.py
code served (permissive licence) · get_code("49077456b34beefd")
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 json
import tempfile
import os

# Create a temporary JSONL file with test data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test.jsonl")

# Write test data to the file
test_data = [
    {"id": 1, "text": "hello world"},
    {"id": 2, "text": "foo bar"},
    {"id": 3, "name": "test", "value": 42}
]

with open(temp_file, 'w') as f:
    for item in test_data:
        json.dump(item, f)
        f.write('\n')

# Define ARGS
ARGS = {
    "fp": temp_file
}
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 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

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 distinct functions with an identity, 3 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 3 later) and 0 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
annotations_from_jsonl 3 other papers
3 later
2005.00115 (2020-05)
Learning to Faithfully Rationalize by Construction
2310.16316 (2023-10)
Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groups
2024.findings-acl.307 (2024 (year only))
Plausible Extractive Rationalization through Semi-Supervised Entailment Signal
load_datasets 3 other papers
3 later
2005.00115 (2020-05)
Learning to Faithfully Rationalize by Construction
2310.16316 (2023-10)
Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groups
2024.findings-acl.307 (2024 (year only))
Plausible Extractive Rationalization through Semi-Supervised Entailment Signal
load_jsonl 3 other papers
3 later
2005.00115 (2020-05)
Learning to Faithfully Rationalize by Construction
2310.16316 (2023-10)
Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groups
2024.findings-acl.307 (2024 (year only))
Plausible Extractive Rationalization through Semi-Supervised Entailment Signal

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

State-of-the-art models in NLP are now predominantly based on deep neural networks that are opaque in terms of how they come to make predictions. This limitation has increased interest in designing more interpretable deep models for NLP that reveal the `reasoning' behind model outputs. But work in this direction has been conducted on different datasets and tasks with correspondingly unique aims and metrics; this makes it difficult to track progress. We propose the Evaluating Rationales And Simple English Reasoning (ERASER) benchmark to advance research on interpretable models in NLP. This benchmark comprises multiple datasets and tasks for which human annotations of "rationales" (supporting evidence) have been collected. We propose several metrics that aim to capture how well the rationales provided by models align with human rationales, and also how faithful these rationales are (i.e., the degree to which provided rationales influenced the corresponding predictions). Our hope is that releasing this benchmark facilitates progress on designing more interpretable NLP systems. The benchmark, code, and documentation are available at https://www.eraserbenchmark.com/

For agents

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

get_harvested_code_for_paper("1911.03429")
get_code_for_paper("1911.03429")
have("1911.03429")

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

curl https://syntology.ai/api/ran/1911.03429.json

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

[![Syntology run record](https://syntology.ai/api/ran/1911.03429.svg)](https://syntology.ai/paper/1911.03429)

Connect an agent — have() is free.