instance_id stringclasses 62
values | model_name stringclasses 3
values | target bool 1
class | trajectory stringlengths 18.4k 147k | exit_status stringclasses 2
values | generated_patch stringlengths 328 9.06k | eval_logs stringlengths 1.65k 36.7k |
|---|---|---|---|---|---|---|
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | [{"cutoff_date": "01.01.2023", "mask": false, "role": "system", "system_prompt": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical ... | submitted |
diff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py
index 7e3938c..d85460a 100644
--- a/cognitive_complexity/utils/ast.py
+++ b/cognitive_complexity/utils/ast.py
@@ -85,7 +85,7 @@ def process_node_itself(
return increment_by, 0, True
elif isinstance(node, ast.BoolOp):
... | Will execute following command for apply_patch:
```
cd //cognitive_complexity &&
source activate base &&
conda activate cognitive_complexity &&
git apply -v /swebench_matterhorn/prediction_patch.patch --allow-empty
```
Checking patch cognitive_complexity/utils/ast.py...
Checking patch tests/test_cognitive_complexity.... |
Melevir__cognitive_complexity-15 | swe-agent-llama-8b | true | [{"cutoff_date": "01.01.2023", "mask": false, "role": "system", "system_prompt": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical ... | submitted |
diff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py
index 7e3938c..61b07f3 100644
--- a/cognitive_complexity/utils/ast.py
+++ b/cognitive_complexity/utils/ast.py
@@ -84,8 +84,10 @@ def process_node_itself(
increment_by += 1
return increment_by, 0, True
elif isinsta... | Will execute following command for apply_patch:
```
cd //cognitive_complexity &&
source activate base &&
conda activate cognitive_complexity &&
git apply -v /swebench_matterhorn/prediction_patch.patch --allow-empty
```
Checking patch cognitive_complexity/utils/ast.py...
Applied patch cognitive_complexity/utils/ast.py... |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted (exit_context) | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
Melevir__cognitive_complexity-15 | swe-agent-llama-70b | true | "[{\"cutoff_date\": \"01.01.2023\", \"mask\": false, \"role\": \"system\", \"system_prompt\": \"SETT(...TRUNCATED) | submitted | "\ndiff --git a/cognitive_complexity/utils/ast.py b/cognitive_complexity/utils/ast.py\nindex 7e3938c(...TRUNCATED) | "Will execute following command for apply_patch:\n```\n\ncd //cognitive_complexity &&\nsource activa(...TRUNCATED) |
End of preview. Expand in Data Studio
Raw SWE-Agent Trajectories
This dataset contains raw software engineering agent trajectories filtered/streamed from:
nebius/SWE-agent-trajectories
Mode:
successful trajectories only- maximum requested samples:
1000 - actual written samples:
1000 - Parquet shards:
20
Important format note
The trajectory column is stored as a JSON string.
That means each row contains:
sample["trajectory"]
as a string like:
[
{
"cutoff_date": "01.01.2023",
"mask": false,
"role": "system",
"system_prompt": "...",
"text": null
},
{
"cutoff_date": null,
"mask": false,
"role": "user",
"system_prompt": null,
"text": "..."
}
]
To parse it:
import json
messages = json.loads(sample["trajectory"])
This raw-string format is used to avoid Arrow nested-list schema problems and to keep the data closer to the original trajectory.
Fields
| Field | Type | Description |
|---|---|---|
instance_id |
string | Repository issue identifier |
model_name |
string | Model used to generate the trajectory |
target |
bool | Whether the issue was solved |
trajectory |
string | Raw JSON string containing full agent conversation |
exit_status |
string | Why the agent stopped |
generated_patch |
string | Final Git patch |
eval_logs |
string | Test execution logs |
Usage
from datasets import load_dataset
import json
ds = load_dataset(
"antontuzovAI/swe-agent-successful-raw-1000",
streaming=True,
split="train"
)
for sample in ds:
print(sample["instance_id"])
print(sample["model_name"])
print(sample["target"])
print(sample["exit_status"])
messages = json.loads(sample["trajectory"])
print("trajectory messages:", len(messages))
break
Source
This dataset is derived from:
nebius/SWE-agent-trajectories
The source dataset contains 80,036 trajectories:
- 13,389 successful trajectories
- 66,647 unsuccessful trajectories
License
The source dataset is licensed under CC-BY-4.0.
However, individual repositories used in SWE-bench instances may have their own licenses.
Also, if you use outputs from Llama-based agents, you must comply with the relevant Llama license.
- Downloads last month
- 87