Dataset Viewer
Auto-converted to Parquet Duplicate
dataset_id
stringclasses
1 value
dataset_title
stringclasses
1 value
source_type
stringclasses
1 value
generated_at
timestamp[ns, tz=UTC]date
2026-06-02 13:04:55
2026-06-02 13:04:55
prototype_count
int64
300
300
prototypes
dict
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{ "_merge_meta": { "candidate_labels": [ "kotlin" ], "stargazers_count": 57305 }, "chunks": [ { "artifact_ref": "release:2.0.1", "artifact_type": "release_note", "memory_node_id": "2dust__v2rayNG__release__2_0_1", "memory_unit_text": "In the 2dust/v2rayNG release 2.0....
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{ "_merge_meta": { "candidate_labels": [], "stargazers_count": null }, "chunks": [ { "artifact_ref": "release:2025.02.19", "artifact_type": "release_note", "memory_node_id": "yt_dlp__yt_dlp__release__2025_02_19", "memory_unit_text": "In the yt-dlp/yt-dlp release 2025.02.19, cor...
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{ "_merge_meta": { "candidate_labels": [ "cli" ], "stargazers_count": 45676 }, "chunks": [ { "artifact_ref": "release:v0.57.0", "artifact_type": "release_note", "memory_node_id": "Aider_AI__aider__release__v0_57_0", "memory_unit_text": "In the Aider-AI/aider release v...
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{ "_merge_meta": { "candidate_labels": [ "swift" ], "stargazers_count": 42391 }, "chunks": [ { "artifact_ref": "release:5.0.5", "artifact_type": "release_note", "memory_node_id": "Alamofire__Alamofire__release__5_0_5", "memory_unit_text": "In the Alamofire/Alamofire r...
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{ "_merge_meta": { "candidate_labels": [ "go" ], "stargazers_count": 49575 }, "chunks": [ { "artifact_ref": "release:v3.38.0", "artifact_type": "release_note", "memory_node_id": "AlistGo__alist__release__v3_38_0", "memory_unit_text": "In the AlistGo/alist release v3.3...
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{"_merge_meta":{"candidate_labels":["browser","mobile"],"stargazers_count":30901},"chunks":[{"artifa(...TRUNCATED)
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{"_merge_meta":{"candidate_labels":["java"],"stargazers_count":33660},"chunks":[{"artifact_ref":"rel(...TRUNCATED)
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{"_merge_meta":{"candidate_labels":["cli","rust"],"stargazers_count":64529},"chunks":[{"artifact_ref(...TRUNCATED)
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{"_merge_meta":{"candidate_labels":["typescript"],"stargazers_count":88161},"chunks":[{"artifact_ref(...TRUNCATED)
official_300repo_release_unified_v1
Official 300-repository GitHub release-note unified benchmark
github_release_note_unified
2026-06-02T13:04:55.376000
300
{"_merge_meta":{"candidate_labels":["browser"],"stargazers_count":42581},"chunks":[{"artifact_ref":"(...TRUNCATED)
End of preview. Expand in Data Studio

TSSMRBench Official 300 Public Release

Dataset Summary

This repository contains the public release package for TSSMRBench, a benchmark for evaluating whether an agent memory system can retrieve the stage-correct temporal semantic state from an evolving memory trajectory.

The source material for this release is derived from public release histories of 300 influential GitHub repositories. For public redistribution, the package removes the original raw_text release-note content and retains only the normalized memory_unit_text representations, benchmark structure, question annotations, and source links.

The release is distributed as a single merged public file:

  • data/official_300_merged_public.json: a single-file public merged release with the same scenario structure as the internal merged dataset, but with raw_text removed.

What TSSMRBench Evaluates

TSSMRBench focuses on temporal semantic state memory retrieval. Each scenario is an evolving memory trajectory composed of versioned semantic states. The benchmark asks whether a memory system can retrieve the state evidence required by a query when several nearby versions are semantically related but temporally different.

The benchmark contains three task families:

  • single_state_lookup: retrieve one stage-correct state.
  • cross_version_comparison: jointly retrieve two states and compare them.
  • temporal_version_ordering: recover several states and place them in the correct temporal progression.

Public Release Design

This Hugging Face release is intended to be easier to redistribute and audit than the internal experiment package:

  • Original raw_text fields are removed.
  • Local absolute paths are removed.
  • source_url links are preserved for traceability.
  • memory_unit_text fields retain the normalized textual state descriptions used in the benchmark.

Data Fields

The merged file official_300_merged_public.json stores one scenario record per repository. Each scenario contains:

  • prototype_id: scenario identifier.
  • repo: source repository.
  • window_title: scenario title.
  • window_summary: summary of the scenario's version history.
  • chunks: the 30 temporal semantic state nodes, each with fields such as memory_node_id, artifact_ref, time_hint, published_at, source_url, and memory_unit_text.
  • questions: the benchmark questions associated with the scenario, including question_id, task_type, query_text, options, correct_option_id, expected_answer, source_chunk_ids, and answer_support.

Intended Use

This release is intended for:

  • benchmarking agent memory systems;
  • retrieval and reranking analysis on evolving memories;
  • error analysis on version-sensitive memory retrieval;
  • research on temporal reasoning grounded in retrieved memory evidence.

Limitations

  • The package is derived from public repository release histories and therefore reflects the style, granularity, and coverage of those sources.
  • memory_unit_text is a normalized rewrite of source material rather than the raw release note text.
  • This public package is designed for benchmark use and traceable redistribution, not as a substitute for the original upstream release documentation.

Licensing Note

This dataset package uses license: other because it combines:

  • benchmark structure, task design, and annotations created by the authors; and
  • derived state descriptions and source links originating from multiple upstream repositories with non-uniform licenses.

Please read the LICENSE file in this repository before redistribution or reuse.

Citation

If you use this dataset, please cite the TSSMRBench paper and repository once the final bibliographic entry is available. Until then, please reference the repository URL directly.

Downloads last month
42