Dataset Viewer
Auto-converted to Parquet Duplicate
instance_id
stringlengths
18
51
coding_id
stringclasses
3 values
coding_model
stringclasses
3 values
repository
stringlengths
7
33
language
stringclasses
5 values
base_commit
stringlengths
40
40
verifier_status
stringclasses
3 values
reference_count
int64
0
9
id
stringlengths
24
58
traj
stringlengths
214k
2.03M
review_comments
stringlengths
276
13.9k
patch
stringlengths
11.9k
135k
reference_patch
stringlengths
12.7k
53.7k
task
stringlengths
571
5.48k
abs-module-cache-flags
gemini
Gemini 3.8 Flash
abs-lang/abs
go
cb1b3b671d0ee9fa9da9f7b02f86967953ffd10a
passed
6
gemini/abs-module-cache-flags
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"c2940642-b5fe-4604-ac6a-3c58ccdeba28\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"abs-module-c(...TRUNCATED)
"diff --git a/evaluator/functions.go b/evaluator/functions.go\nindex 25cdbd1..b841430 100644\n--- a/(...TRUNCATED)
"diff --git a/evaluator/functions.go b/evaluator/functions.go\nindex 25cdbd1..9ddc8ed 100644\n--- a/(...TRUNCATED)
"Improve ABS module loading so `require()` remains deterministic across larger dependency graphs, su(...TRUNCATED)
abs-stepped-slices
gemini
Gemini 3.8 Flash
abs-lang/abs
go
cb1b3b671d0ee9fa9da9f7b02f86967953ffd10a
passed
4
gemini/abs-stepped-slices
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"b88a705e-b3b5-4f37-bfb6-2fd59fc7b14d\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"abs-stepped-(...TRUNCATED)
"diff --git a/ast/ast.go b/ast/ast.go\nindex 998634f..dd33fec 100644\n--- a/ast/ast.go\n+++ b/ast/as(...TRUNCATED)
"diff --git a/ast/ast.go b/ast/ast.go\nindex 998634f..8a6fec6 100644\n--- a/ast/ast.go\n+++ b/ast/as(...TRUNCATED)
"Extend indexing ranges so arrays and strings support a third slice component:\n\n- `value[start:end(...TRUNCATED)
actionlint-action-pinning-lint
gemini
Gemini 3.8 Flash
rhysd/actionlint
go
0bdc95715fa58f64e3fd6e63b0f89be8733cbbab
passed
3
gemini/actionlint-action-pinning-lint
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"c2f86ede-3c78-4e64-ac63-47cea1407bd7\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"actionlint-a(...TRUNCATED)
"diff --git a/command.go b/command.go\nindex b68784e..d614ad4 100644\n--- a/command.go\n+++ b/comman(...TRUNCATED)
"diff --git a/command.go b/command.go\nindex b68784e..1cdfc20 100644\n--- a/command.go\n+++ b/comman(...TRUNCATED)
"Teams need to enforce that action and reusable workflow references use pinned versions rather than (...TRUNCATED)
adaptix-name-mapping-aliases
gemini
Gemini 3.8 Flash
reagento/adaptix
python
a691069fcadf9131e5f7a5a130a022dc678f3e1d
passed
1
gemini/adaptix-name-mapping-aliases
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"a7521142-b37b-44ce-9a39-e83ea6007c45\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"adaptix-name(...TRUNCATED)
"diff --git a/src/adaptix/_internal/morphing/facade/provider.py b/src/adaptix/_internal/morphing/fac(...TRUNCATED)
"diff --git a/src/adaptix/_internal/morphing/facade/provider.py b/src/adaptix/_internal/morphing/fac(...TRUNCATED)
"`name_mapping` can rename fields via `map` but cannot accept multiple alternative input keys for th(...TRUNCATED)
aiomonitor-task-snapshots-diff
gemini
Gemini 3.8 Flash
aio-libs/aiomonitor
python
b73fea2e0682803bda7531c93cd1dfb360839175
passed
1
gemini/aiomonitor-task-snapshots-diff
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"6eadcee9-39ca-44c0-a999-0acabfb8e648\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"aiomonitor-t(...TRUNCATED)
"diff --git a/aiomonitor/monitor.py b/aiomonitor/monitor.py\nindex 5a2ad70..dfedf64 100644\n--- a/ai(...TRUNCATED)
"diff --git a/aiomonitor/monitor.py b/aiomonitor/monitor.py\nindex 5a2ad70..6f3d707 100644\n--- a/ai(...TRUNCATED)
"aiomonitor lacks the ability to capture and compare task state over time.\n\nAdd snapshots to Monit(...TRUNCATED)
anko-default-function-arguments
gemini
Gemini 3.8 Flash
mattn/anko
go
9d2d84bb1564e9513287998c56ccf16c01c19008
passed
2
gemini/anko-default-function-arguments
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"67781edb-71c7-43f1-a530-7169321ff7b1\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"anko-default(...TRUNCATED)
"diff --git a/ast/astutil/walk.go b/ast/astutil/walk.go\nindex 19068cb..bd5b4b3 100644\n--- a/ast/as(...TRUNCATED)
"diff --git a/core/core.go b/core/core.go\nindex 2699257..2d27525 100644\n--- a/core/core.go\n+++ b/(...TRUNCATED)
"Add support for default argument values written as `name = expression` in function parameter lists.(...TRUNCATED)
anko-typed-variable-bindings
gemini
Gemini 3.8 Flash
mattn/anko
go
3f269a72ff69398b1250c584171f32d12c0d8085
passed
0
gemini/anko-typed-variable-bindings
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"95b4ff8f-1fe5-445b-a530-21bbf740754c\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"anko-typed-v(...TRUNCATED)
"diff --git a/ast/stmt.go b/ast/stmt.go\nindex e7a15a8..c7d6697 100644\n--- a/ast/stmt.go\n+++ b/ast(...TRUNCATED)
"diff --git a/ast/stmt.go b/ast/stmt.go\nindex e7a15a8..5bd6532 100644\n--- a/ast/stmt.go\n+++ b/ast(...TRUNCATED)
"Anko variables are dynamically typed, with no mechanism to enforce type constraints after declarati(...TRUNCATED)
arcane-drift-detection-baselines
gemini
Gemini 3.8 Flash
getarcaneapp/arcane
go
d34a5e2a6c5eb0f0955039775f5b9538424b58ff
passed
3
gemini/arcane-drift-detection-baselines
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"8ab6ad20-2304-4931-b54b-c26eef08e0ea\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"arcane-drift(...TRUNCATED)
"diff --git a/backend/internal/bootstrap/jobs_bootstrap.go b/backend/internal/bootstrap/jobs_bootstr(...TRUNCATED)
"diff --git a/backend/internal/bootstrap/jobs_bootstrap.go b/backend/internal/bootstrap/jobs_bootstr(...TRUNCATED)
"Implement a drift detection engine comparing live container state against baselines. Follow pattern(...TRUNCATED)
arktype-json-schema-refs-dependencies
gemini
Gemini 3.8 Flash
arktypeio/arktype
typescript
04355e8b26d1ad5264ef62314a2bc46c4de58ed8
failed
4
gemini/arktype-json-schema-refs-dependencies
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"fd88c9c9-3662-431d-834d-7bae74c2c6e4\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"arktype-json(...TRUNCATED)
"diff --git a/ark/json-schema/__tests__/common.test.ts b/ark/json-schema/__tests__/common.test.ts\nn(...TRUNCATED)
"diff --git a/ark/json-schema/array.ts b/ark/json-schema/array.ts\nindex 4455638b..dc848ba7 100644\n(...TRUNCATED)
"Expected Feature:\ndependencies/dependentRequired: if trigger key present, require dependent keys.\(...TRUNCATED)
awilix-async-container-initialization
gemini
Gemini 3.8 Flash
jeffijoe/awilix
typescript
82ac179c1de4c216c4e333093044fac643303f0c
failed
4
gemini/awilix-async-container-initialization
"{\n \"schema_version\": \"ATIF-v1.7\",\n \"session_id\": \"b47fdaea-9f5f-48ac-8020-bc86541ed7d9\"(...TRUNCATED)
"{\n \"schema_version\": \"stage2-pilot17-reference-review-v1\",\n \"instance_id\": \"awilix-async(...TRUNCATED)
"diff --git a/src/__tests__/container.initialize.test.ts b/src/__tests__/container.initialize.test.t(...TRUNCATED)
"diff --git a/src/awilix.ts b/src/awilix.ts\nindex e43989b..9fd9fec 100644\n--- a/src/awilix.ts\n+++(...TRUNCATED)
"Add support for asynchronous initialization of container registrations with automatic dependency-aw(...TRUNCATED)
End of preview. Expand in Data Studio

Diffs Are Cheap, Show Me the Trajectory

Process-Aware Code Review
Coding trajectories, synthetic reference comments, and implementation patches.

Quick start   /   Data overview   /   Data fields   /   Code and experiments

What is included

113 DeepSWE tasks, three coding models, 339 complete coding trajectories, and 882 synthetic reference comments. Each record pairs a coding trajectory with its implementation patch and reference review comments.

Core data Contents Field
Coding trajectories Complete ATIF-v1.7 development records, including agent messages and tool interactions traj
Reference comments Synthetic reference findings with original descriptions, evidence, and metadata review_comments
Implementation patches Coding-agent changes relative to the upstream base commit patch
Task context Task description, upstream repository, programming language, and base commit task, repository, language, base_commit
Reference solutions Held-out reference diffs, separate from coding-agent implementations reference_patch
Verification outcomes Recorded pass, failure, or harness timeout verifier_status

All 339 coding-agent trajectories are complete, not summaries or truncated exports. Each coding model implements the same 113 tasks; these are 339 model-task records, not 339 distinct tasks.

Not included: standalone benchmark task definitions, task Dockerfiles, hidden test patches, repository snapshots, reviewer execution traces, or the full experimental reproduction code. See the source repository for the broader experimental artifact.

Find the coding trajectories

Configuration Coding model Records Reference comments Data
gpt55 GPT-5.5 113 355 JSONL
gpt56 GPT-5.6 Sol 113 185 JSONL
gemini Gemini 3.8 Flash 113 342 JSONL
all All three models 339 882 All files above

The default all configuration combines the three coding cohorts. Model configurations are not independent train/test splits. The single train split is a storage convention, not a prescribed evaluation protocol.

Quick start

pip install datasets
import json
from datasets import load_dataset

data = load_dataset("ZijiantheJack/ProcessAwareCodeReview", "all", split="train")
example = data[0]

trajectory = json.loads(example["traj"])
reference_review = json.loads(example["review_comments"])
comments = reference_review["reviews"]
implementation_patch = example["patch"]

Select gpt55, gpt56, or gemini instead of all to load one coding model. Use instance_id to align the three implementations of the same task.

Data fields

Field Contents
id Unique coding_id/instance_id key
instance_id Task identifier shared across the three coding models
coding_id, coding_model Coding configuration and model name
repository, base_commit, language Upstream project context
traj Complete original ATIF-v1.7 JSON text; parse with json.loads
review_comments Complete original reference-review JSON text, including reviews and validation metadata
patch Original coding-agent implementation diff
reference_patch Separate reference solution diff
task Task description supplied to the coding agent
reference_count Number of comments in the reference record
verifier_status Recorded official outcome: passed, failed, or harness_timeout

JSON-valued fields are UTF-8 strings so heterogeneous native ATIF steps and reference schemas load without flattening, truncation, or schema inference loss. After JSONL decoding, these strings and the patch/task fields preserve the original source text exactly.

Each reference comment retains its original ID, title, description, expected behavior, severity, file, region, failure scenario, supporting evidence, and confidence. Empty comment lists are retained: 2 GPT-5.5, 21 GPT-5.6 Sol, and 3 Gemini records have no reference comments. No implementation patch is empty.

Synthetic reference comments

Reference comments are model-generated reference labels, not human reviews or outputs from the different review methods in the broader experiment. Reference generation uses task text and implementation/reference patches, without consuming the coding trajectory.

Original metadata are retained, including reference_kind, validation_status, and deduplication_status. Records include synthetic_unvalidated_signal_closure with validation and deduplication skipped; a reference label is not a guarantee of a verified defect. The original comment severity is not a newly inferred failure-linked label.

Trajectories contain original agent messages, tool inputs and outputs, code excerpts, generated tests, and recorded local paths. Model-produced content and findings can be incorrect.

Verification outcomes

Coding model Passed Failed Harness timeout
GPT-5.5 62 51 0
GPT-5.6 Sol 82 31 0
Gemini 3.8 Flash 84 28 1
Total 228 110 1

These are recorded outcomes, not newly executed verification. A timeout is not silently treated as a pass or failure.

Layout

data/
  gpt55/train.jsonl
  gpt56/train.jsonl
  gemini/train.jsonl
assets/
  logo.svg
provenance/
  LICENSES/
  artifact/origin.json
  docs/PROVENANCE.md
  export-report.json
  source-paths.json
SHA256SUMS.json

Rights and attribution

The source artifact does not grant a blanket license for all included content. DeepSWE's Apache-2.0 notice covers its original contributions and does not relicense upstream code. Task patches remain subject to their upstream projects' terms; the source framework's ISC declaration does not relicense benchmark data.

Retained notices: DeepSWE Apache-2.0 and upstream attribution. Consult those records and the relevant upstream terms before redistribution or reuse. This dataset does not assert new licensing rights.

Downloads last month
65