Upload HERO-Trajectories release

.gitattributes

@@ -58,3 +58,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+verified/HERO-35B-A3B/results.jsonl filter=lfs diff=lfs merge=lfs -text
+verified/HERO-4B/results.jsonl filter=lfs diff=lfs merge=lfs -text
+verified/HERO-9B/results.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,105 @@
 ---
-license: apache-2.0
+language:
+- en
+pretty_name: HERO Evaluation Trajectories
+size_categories:
+- 1K<n<10K
+tags:
+- agent-trajectories
+- coding-agents
+- swe-bench
+- evaluation
 ---
+
+<p align="center">
+  <img src="assets/hero-logo.png" alt="HERO logo" width="760">
+</p>
+
+<h1 align="center">HERO Evaluation Trajectories</h1>
+
+<p align="center">
+  <strong>Doing More with Less Tokens: Hierarchical Reinforcement Learning for Efficient Coding Agents</strong>
+</p>
+
+<p align="center">
+  <a href="https://arxiv.org/abs/2609.38885"><img src="https://img.shields.io/badge/arXiv-2609.38885-b31b1b.svg" alt="Paper"></a>
+  <a href="https://github.com/XLearning-SCU/HERO"><img src="https://img.shields.io/badge/GitHub-Code-181717.svg" alt="Code"></a>
+  <a href="https://huggingface.co/collections/XLearning-SCU/hero"><img src="https://img.shields.io/badge/%F0%9F%A4%97-Models-f4c430.svg" alt="Hugging Face Models"></a>
+  <a href="https://huggingface.co/datasets/XLearning-SCU/HERO-Trajectories"><img src="https://img.shields.io/badge/%F0%9F%A4%97-Trajectories-f4c430.svg" alt="Hugging Face Trajectories"></a>
+</p>
+
+## 🔍 Overview
+
+HERO is a hierarchical reinforcement learning method for training token-efficient coding agents while prioritizing task resolution. This release contains 2,400 evaluation trajectories from HERO-4B, HERO-9B, and HERO-35B-A3B on SWE-bench Verified and SWE-bench Multilingual, using Claude Code as the agent scaffold.
+
+## Results
+
+Each model includes 500 Verified tasks and 300 Multilingual tasks. These are individual recorded runs; the paper reports averages over three random seeds.
+
+| Benchmark | Model | Tasks | Resolved Rate | Token Usage (M) |
+| --- | --- | ---: | ---: | ---: |
+| Verified | [HERO-4B](verified/HERO-4B/) | 500 | 43.2% | 3.5 |
+| Verified | [HERO-9B](verified/HERO-9B/) | 500 | 50.6% | 2.8 |
+| Verified | [HERO-35B-A3B](verified/HERO-35B-A3B/) | 500 | 61.4% | 2.4 |
+| Multilingual | [HERO-4B](multilingual/HERO-4B/) | 300 | 29.3% | 4.2 |
+| Multilingual | [HERO-9B](multilingual/HERO-9B/) | 300 | 40.0% | 3.6 |
+| Multilingual | [HERO-35B-A3B](multilingual/HERO-35B-A3B/) | 300 | 46.0% | 2.7 |
+
+Tokens are averages per task in millions (M), summed over agent calls, including repeated input context. Records with missing or zero usage are excluded from token averages; resolve rates include all tasks. Detailed usage and processing notes are in each `summary.json`.
+
+## Files
+
+Each `<benchmark>/<model>/` directory contains:
+
+| File | Contents |
+| --- | --- |
+| `results.jsonl` | Task outcomes, grading reports, token usage, and trajectory paths |
+| `predictions.jsonl` | Predicted patches |
+| `summary.json` | Aggregate results, evaluation settings, and processing notes |
+| `trajectories.tar.gz` | Per-task JSON trajectories in `hero-trace-v1` format |
+
+## Loading the Traces
+
+Extract an archive from the repository root:
+
+```bash
+tar -xzf verified/HERO-4B/trajectories.tar.gz -C verified/HERO-4B
+```
+
+Load an extracted trajectory:
+
+```python
+import json
+from pathlib import Path
+
+path = next(Path("verified/HERO-4B/trajectories").glob("*.json"))
+with path.open() as f:
+    trace = json.load(f)
+calls = trace["trajectory"]
+```
+
+The `trajectory_file` paths in `results.jsonl` are relative to the model directory.
+
+## Run Notes
+
+The existing 35B run summaries report different checkpoints for Verified and Multilingual; checkpoint correspondence with the released weights remains unconfirmed. Some prediction patches were sanitized without re-evaluation, so resolution labels describe the original runs. The 35B Multilingual token statistics exclude auxiliary token-counting calls, as documented in its `summary.json`.
+
+## Source Licenses
+
+[SWE-bench Verified](https://huggingface.co/datasets/princeton-nlp/SWE-bench_Verified), [SWE-bench Multilingual](https://huggingface.co/datasets/SWE-bench/SWE-bench_Multilingual), and their source repositories retain their respective licenses. Please also cite the corresponding benchmarks when using these traces.
+
+## 📖 Citation
+
+If you find HERO useful, please cite our [paper](https://arxiv.org/abs/2609.38885):
+
+```bibtex
+@misc{li2026hero,
+  title         = {Doing More with Less Tokens: Hierarchical Reinforcement Learning for Efficient Coding Agents},
+  author        = {Haobin Li and Liang Jiang and Zhenyu Huang and Mouxing Yang and Xi Peng},
+  year          = {2026},
+  eprint        = {2609.38885},
+  archivePrefix = {arXiv},
+  primaryClass  = {cs.SE},
+  url           = {https://arxiv.org/abs/2609.38885}
+}
+```

multilingual/HERO-35B-A3B/summary.json

@@ -0,0 +1,49 @@
+{
+  "schema_version": "hero-trace-v1",
+  "model": "HERO-35B-A3B",
+  "dataset": "SWE-bench Multilingual",
+  "dataset_file_sha256": "f2abb829012990b8b70f0a05414258d285515dd3fd81260a933d71e335e8ffc6",
+  "dataset_revision": "Upstream revision not recorded; dataset_file_sha256 identifies the evaluated input snapshot.",
+  "total": 300,
+  "resolved": 138,
+  "resolved_rate": 0.46,
+  "valid_usage_count": 293,
+  "average_total_tokens_valid_usage": 2714171.0307167238,
+  "average_input_tokens_valid_usage": 2700824.928327645,
+  "average_output_tokens_valid_usage": 13346.102389078498,
+  "token_average_definition": "Exclude records with metrics_missing=true or total_tokens<=0; resolution rate includes every instance. Token usage excludes verified, unreferenced count_tokens auxiliary generations from historical runs; retained calls keep their original recorded usage (including repeated context). This is post-processing, not a rerun.",
+  "evaluation_config": {
+    "TEMPERATURE": "0.6"
+  },
+  "postprocessing": {
+    "kind": "historical_count_tokens_cleanup",
+    "date": "2026-09-21",
+    "excluded_call_count": 140,
+    "affected_instance_count": 101,
+    "retained_candidate_count": 0,
+    "excluded_input_tokens": 1976607,
+    "excluded_output_tokens": 176052,
+    "excluded_total_tokens": 2152659,
+    "equivalent_to_rerun": false
+  },
+  "release_sanitization": {
+    "model_names": "Public model-family names; private checkpoint identifiers removed.",
+    "environment": "Host identifiers, personal home prefixes, deployment endpoints and runtime environment values anonymized. Generic sandbox paths and public project test fixtures remain.",
+    "prediction_patches_sanitized": 1,
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "reevaluated_after_sanitization": false
+  },
+  "evaluation_provenance": "Verified and Multilingual were evaluated with different checkpoints of this model family. These results must not be represented as a single checkpoint evaluated on both datasets. Exact correspondence to separately distributed weights has not been verified.",
+  "average_turns_valid_usage": 46.6382252559727,
+  "status_counts": {
+    "COMPLETED": 264,
+    "TIMEOUT": 7,
+    "ERROR": 29
+  },
+  "recorded_usage_sums_valid_usage": {
+    "input_tokens": 791341704,
+    "output_tokens": 3910408,
+    "total_tokens": 795252112
+  }
+}

multilingual/HERO-35B-A3B/trajectories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:397fd774e5fa72d017ea8d26f42161e0076b516d37112f16a74a6ae89dc332d6
+size 762360169

multilingual/HERO-4B/summary.json

@@ -0,0 +1,44 @@
+{
+  "schema_version": "hero-trace-v1",
+  "model": "HERO-4B",
+  "dataset": "SWE-bench Multilingual",
+  "dataset_file_sha256": "9bd3f0651fd07eaf6442f463ae6f08ed51dd0044b209ff075faa094983daeec7",
+  "dataset_revision": "Upstream revision not recorded; dataset_file_sha256 identifies the evaluated input snapshot.",
+  "total": 300,
+  "resolved": 88,
+  "resolved_rate": 0.29333333333333333,
+  "valid_usage_count": 300,
+  "average_total_tokens_valid_usage": 4243285.973333334,
+  "average_input_tokens_valid_usage": 4219874.3,
+  "average_output_tokens_valid_usage": 23411.673333333332,
+  "token_average_definition": "Recorded per-call input+output, including repeated context. No auxiliary-call filtering or retokenization. Exclude missing/zero usage; resolution includes every instance.",
+  "evaluation_config": {
+    "EVAL_TEST_TIMEOUT": "1500",
+    "TASK_TIMEOUT": "5100"
+  },
+  "postprocessing": {
+    "kind": "format_conversion",
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "patches_changed": true
+  },
+  "release_sanitization": {
+    "model_names": "Public model-family names; private checkpoint identifiers removed.",
+    "environment": "Host identifiers, personal home prefixes, deployment endpoints and runtime environment values anonymized. Generic sandbox paths and public project test fixtures remain.",
+    "prediction_patches_sanitized": 3,
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "reevaluated_after_sanitization": false
+  },
+  "average_turns_valid_usage": 65.02333333333333,
+  "status_counts": {
+    "COMPLETED": 183,
+    "ERROR": 113,
+    "TIMEOUT": 4
+  },
+  "recorded_usage_sums_valid_usage": {
+    "input_tokens": 1265962290,
+    "output_tokens": 7023502,
+    "total_tokens": 1272985792
+  }
+}

multilingual/HERO-4B/trajectories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:646705f6ae8fbd492b798350953656360218f130bcff4202604b72407f5920a4
+size 1192076319

multilingual/HERO-9B/summary.json

@@ -0,0 +1,44 @@
+{
+  "schema_version": "hero-trace-v1",
+  "model": "HERO-9B",
+  "dataset": "SWE-bench Multilingual",
+  "dataset_file_sha256": "9bd3f0651fd07eaf6442f463ae6f08ed51dd0044b209ff075faa094983daeec7",
+  "dataset_revision": "Upstream revision not recorded; dataset_file_sha256 identifies the evaluated input snapshot.",
+  "total": 300,
+  "resolved": 120,
+  "resolved_rate": 0.4,
+  "valid_usage_count": 300,
+  "average_total_tokens_valid_usage": 3646165.3133333335,
+  "average_input_tokens_valid_usage": 3622375.486666667,
+  "average_output_tokens_valid_usage": 23789.826666666668,
+  "token_average_definition": "Recorded per-call input+output, including repeated context. No auxiliary-call filtering or retokenization. Exclude missing/zero usage; resolution includes every instance.",
+  "evaluation_config": {
+    "EVAL_TEST_TIMEOUT": "1500",
+    "TASK_TIMEOUT": "5100"
+  },
+  "postprocessing": {
+    "kind": "format_conversion",
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "patches_changed": true
+  },
+  "release_sanitization": {
+    "model_names": "Public model-family names; private checkpoint identifiers removed.",
+    "environment": "Host identifiers, personal home prefixes, deployment endpoints and runtime environment values anonymized. Generic sandbox paths and public project test fixtures remain.",
+    "prediction_patches_sanitized": 1,
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "reevaluated_after_sanitization": false
+  },
+  "average_turns_valid_usage": 60.03666666666667,
+  "status_counts": {
+    "COMPLETED": 212,
+    "ERROR": 79,
+    "TIMEOUT": 9
+  },
+  "recorded_usage_sums_valid_usage": {
+    "input_tokens": 1086712646,
+    "output_tokens": 7136948,
+    "total_tokens": 1093849594
+  }
+}

multilingual/HERO-9B/trajectories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:043f5d4707f43258c4ac6ab96256ba86552576c69ac75607c5ea876eafdf0f42
+size 1034001861

verified/HERO-35B-A3B/results.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61fc387979a7a72fec96e434fdc244346ee5d434e178e8aa578c1b56324d15b4
+size 11660923

verified/HERO-35B-A3B/summary.json

@@ -0,0 +1,46 @@
+{
+  "schema_version": "hero-trace-v1",
+  "model": "HERO-35B-A3B",
+  "dataset": "SWE-bench Verified",
+  "dataset_file_sha256": "9afdb5f9ec06c93c5decfcaffc837658ad65c87bc3d263d857d19fbe05dea889",
+  "dataset_revision": "Upstream revision not recorded; dataset_file_sha256 identifies the evaluated input snapshot.",
+  "total": 500,
+  "resolved": 307,
+  "resolved_rate": 0.614,
+  "valid_usage_count": 500,
+  "average_total_tokens_valid_usage": 2416620.958,
+  "average_input_tokens_valid_usage": 2405621.552,
+  "average_output_tokens_valid_usage": 10999.406,
+  "token_average_definition": "Recorded per-call input+output, including repeated context. No auxiliary-call filtering or retokenization. Exclude missing/zero usage; resolution includes every instance.",
+  "evaluation_config": {
+    "EVAL_TEST_TIMEOUT": "1500",
+    "MAX_CONTEXT_TOKENS": "131072",
+    "MAX_RESPONSE_TOKENS": "16000",
+    "TASK_TIMEOUT": "5100"
+  },
+  "postprocessing": {
+    "kind": "format_conversion",
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "patches_changed": true
+  },
+  "release_sanitization": {
+    "model_names": "Public model-family names; private checkpoint identifiers removed.",
+    "environment": "Host identifiers, personal home prefixes, deployment endpoints and runtime environment values anonymized. Generic sandbox paths and public project test fixtures remain.",
+    "prediction_patches_sanitized": 1,
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "reevaluated_after_sanitization": false
+  },
+  "evaluation_provenance": "Verified and Multilingual were evaluated with different checkpoints of this model family. These results must not be represented as a single checkpoint evaluated on both datasets. Exact correspondence to separately distributed weights has not been verified.",
+  "average_turns_valid_usage": 48.052,
+  "status_counts": {
+    "COMPLETED": 451,
+    "ERROR": 49
+  },
+  "recorded_usage_sums_valid_usage": {
+    "input_tokens": 1202810776,
+    "output_tokens": 5499703,
+    "total_tokens": 1208310479
+  }
+}

verified/HERO-35B-A3B/trajectories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b068887c275d36372625c90b713538fb8b861f671a51124ea056bb15a49c673e
+size 1280256957

verified/HERO-4B/results.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:046dd2d878beb386240f5332aba957c495af3e5168bbd376982452f91d066e88
+size 12780968

verified/HERO-4B/summary.json

@@ -0,0 +1,46 @@
+{
+  "schema_version": "hero-trace-v1",
+  "model": "HERO-4B",
+  "dataset": "SWE-bench Verified",
+  "dataset_file_sha256": "9afdb5f9ec06c93c5decfcaffc837658ad65c87bc3d263d857d19fbe05dea889",
+  "dataset_revision": "Upstream revision not recorded; dataset_file_sha256 identifies the evaluated input snapshot.",
+  "total": 500,
+  "resolved": 216,
+  "resolved_rate": 0.432,
+  "valid_usage_count": 500,
+  "average_total_tokens_valid_usage": 3524970.364,
+  "average_input_tokens_valid_usage": 3506947.46,
+  "average_output_tokens_valid_usage": 18022.904,
+  "token_average_definition": "Recorded per-call input+output, including repeated context. No auxiliary-call filtering or retokenization. Exclude missing/zero usage; resolution includes every instance.",
+  "evaluation_config": {
+    "EVAL_TEST_TIMEOUT": "1500",
+    "MAX_CONTEXT_TOKENS": "131072",
+    "MAX_RESPONSE_TOKENS": "16000",
+    "TASK_TIMEOUT": "5100"
+  },
+  "postprocessing": {
+    "kind": "format_conversion",
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "patches_changed": true
+  },
+  "release_sanitization": {
+    "model_names": "Public model-family names; private checkpoint identifiers removed.",
+    "environment": "Host identifiers, personal home prefixes, deployment endpoints and runtime environment values anonymized. Generic sandbox paths and public project test fixtures remain.",
+    "prediction_patches_sanitized": 3,
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "reevaluated_after_sanitization": false
+  },
+  "average_turns_valid_usage": 54.034,
+  "status_counts": {
+    "COMPLETED": 335,
+    "ERROR": 159,
+    "TIMEOUT": 6
+  },
+  "recorded_usage_sums_valid_usage": {
+    "input_tokens": 1753473730,
+    "output_tokens": 9011452,
+    "total_tokens": 1762485182
+  }
+}

verified/HERO-4B/trajectories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2a37d8163374f048844f501c01d9417dfaefd6a94581f5019423d33585ceeec
+size 1744497399

verified/HERO-9B/results.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9989d450f6908a70c5148f92b7c862c12e1b629aef7bf081e4b5cee49fee04ad
+size 11678368

verified/HERO-9B/summary.json

@@ -0,0 +1,46 @@
+{
+  "schema_version": "hero-trace-v1",
+  "model": "HERO-9B",
+  "dataset": "SWE-bench Verified",
+  "dataset_file_sha256": "9afdb5f9ec06c93c5decfcaffc837658ad65c87bc3d263d857d19fbe05dea889",
+  "dataset_revision": "Upstream revision not recorded; dataset_file_sha256 identifies the evaluated input snapshot.",
+  "total": 500,
+  "resolved": 253,
+  "resolved_rate": 0.506,
+  "valid_usage_count": 500,
+  "average_total_tokens_valid_usage": 2800130.248,
+  "average_input_tokens_valid_usage": 2782844.626,
+  "average_output_tokens_valid_usage": 17285.622,
+  "token_average_definition": "Recorded per-call input+output, including repeated context. No auxiliary-call filtering or retokenization. Exclude missing/zero usage; resolution includes every instance.",
+  "evaluation_config": {
+    "EVAL_TEST_TIMEOUT": "1500",
+    "MAX_CONTEXT_TOKENS": "131072",
+    "MAX_RESPONSE_TOKENS": "16000",
+    "TASK_TIMEOUT": "5100"
+  },
+  "postprocessing": {
+    "kind": "format_conversion",
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "patches_changed": true
+  },
+  "release_sanitization": {
+    "model_names": "Public model-family names; private checkpoint identifiers removed.",
+    "environment": "Host identifiers, personal home prefixes, deployment endpoints and runtime environment values anonymized. Generic sandbox paths and public project test fixtures remain.",
+    "prediction_patches_sanitized": 1,
+    "resolved_labels_changed": false,
+    "token_counts_changed": false,
+    "reevaluated_after_sanitization": false
+  },
+  "average_turns_valid_usage": 47.018,
+  "status_counts": {
+    "COMPLETED": 397,
+    "ERROR": 98,
+    "TIMEOUT": 5
+  },
+  "recorded_usage_sums_valid_usage": {
+    "input_tokens": 1391422313,
+    "output_tokens": 8642811,
+    "total_tokens": 1400065124
+  }
+}

verified/HERO-9B/trajectories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8555ba47ed29b09afcc9b5793a4440e6b26797b6981971e59457b589588bfdf
+size 1339877582