Datasets:

naivecynics
/

MetaScore-SFT-100

+# Additional benchmark details
+## Selection and fixed windows
+The existing selection pipeline retained 106 candidate scores satisfying:
+- A single piano part (`keyboard.piano*`), at least 16 measures, no lyrics.
+- The upstream best-unique-arrangement filter.
+- Rating >= 4.5 with at least 3 ratings.
+- The upstream 4/4 filter; the final MIDI window is checked explicitly against
+  the continuation protocol.
+Manual listening rejected six clear quality-control failures; their CIDs are
+preserved in `summary.json`. The remaining 100 pieces span 15 genre labels.
+These genres are not equally represented; see the summary for exact counts.
+The deterministic window builder accounts for score pickup information,
+merges piano events and duplicate `(start, pitch)` notes, and chooses the first
+complete protocol-compatible 16-bar window. **99 pieces use the opening window;
+one uses the first subsequent compatible window** because the opening does not
+meet the protocol. `challenge_test/manifest.json` records the source SHA256,
+window index, pickup, selection reason, note counts, and MIDI parsing diagnostics.
+No new selection or window extraction was performed for this release.
+## Format
+Times use a sixteenth-note grid. A complete window ends at step 272, with steps
+0--15 reserved for an optional pickup. The four regular prompt bars occupy
+steps 16--79. Prompt notes have start in 0--79, and reference notes in 80--271.
+Pitch is an integer in 0--127; duration is a positive integer and notes do not
+extend beyond step 272. Notes are sorted by `(start, pitch)`.
+```json
+{"prompt": [{"start": 16, "pitch": 60, "duration": 4}]}
+```
+```json
+{"generation": [{"start": 80, "pitch": 64, "duration": 4}]}
+```
+The fixed-window MIDIs use 480 ticks per quarter note, 120 BPM, program 0 and
+velocity 64 for comparable listening. Original complete-piece assets retain
+their original contents. Genre labels and source previews are auxiliary;
+they are not additional generation conditions in the DAMN benchmark.
+## Provenance, permissions and limitations
+Metadata records 99 source scores as `cc-zero` and one as `publicdomain`.
+These are upstream declarations, not an independent verification of rights to
+every underlying composition or arrangement. Existing score credits and copyright
+notices are retained, and the explicit conflicting entry is included only in this private repository, without clearing its rights. There is
+no blanket license replacing source rights; see [LICENSE.md](LICENSE.md).
+The benchmark contains familiar repertoire and a small, curated selection of
+genres. Source metadata can be imperfect. References are one possible
+continuation rather than the only musically valid answer. Similarity to a
+reference is a development proxy, not a substitute for listening evaluation.
+The DAMN pipeline excludes benchmark matches from its training data; this does
+not establish the absence of overlap with other models' training corpora.

README.md CHANGED Viewed

@@ -21,120 +21,65 @@ configs:
 # MetaScore-SFT-100
-A fixed, 100-piece symbolic piano continuation development benchmark derived
-from [MetaScore](https://github.com/wx83/MetaScore_Official), used to evaluate
-[DAMN](https://github.com/naivecynics/MIREX2026-DAMN/tree/mirex2026-submission).
-**This is an independently curated development benchmark, not the official
-MIREX 2026 test set.** It contains one fixed 4-bar prompt (plus an optional pickup)
-and a 12-bar reference continuation for each piece.
-## Release completeness
-This private repository contains all **100 fixed entries**, including complete source assets, prompts and references. The reported evaluation uses this same 100-piece benchmark.
-One source has conflicting CC0 metadata and an embedded "All Rights Reserved" notice. Its content is included for private review; its rights are not reclassified or cleared. Original attribution is retained in `attribution.json` and the conflict is recorded in `rights_conflicts.json`. This repository remains private.
-## Contents
-| Path | Description |
 | --- | --- |
-| `test.jsonl` | 100 Viewer entries with CID, genre, source, availability, and prompt/reference notes |
-| `challenge_test/prompts/` | 100 byte-preserved MIREX-format prompt JSON files |
-| `challenge_test/references/` | 100 byte-preserved reference JSON files |
-| `challenge_test/midi/prompts/` | 100 prompt MIDIs |
-| `challenge_test/midi/reference_full/` | 100 prompt + reference MIDIs |
-| `challenge_test/manifest.json` | Original 100-window manifest, unchanged |
-| `midi/`, `mscz/`, `mp3/` | 100 complete source MIDIs, MuseScore scores, and existing source preview MP3s per directory |
-| `label/` | 100 original genre-label JSON files |
-| `manifest.csv`, `summary.json` | Original selection metadata and filtering summary for all 100 pieces |
-| `attribution.json` | Per-piece source links, declared licenses, embedded score credits and release status |
-| `rights_conflicts.json` | Preserved source-license conflict and affected asset checksums |
-| `release_summary.json` | Release counts and original challenge-manifest checksum |
-| `checksums.sha256` | SHA256 of every published file except the checksum file itself |
-| `RECONSTRUCTION.md` | How the fixed benchmark was built |
-All included raw assets and original test files are copied without modification.
-The original MP3s are full-source previews; they are not newly rendered versions
-of the fixed 16-bar evaluation windows. Source metadata spelling is retained.
-## Selection and fixed windows
-The existing selection pipeline retained 106 candidate scores satisfying:
-- A single piano part (`keyboard.piano*`), at least 16 measures, no lyrics.
 - The upstream best-unique-arrangement filter.
-- Rating >= 4.5 with at least 3 ratings.
-- The upstream 4/4 filter; the final MIDI window is checked explicitly against
-  the continuation protocol.
-Manual listening rejected six clear quality-control failures; their CIDs are
-preserved in `summary.json`. The remaining 100 pieces span 15 genre labels.
-These genres are not equally represented; see the summary for exact counts.
-The deterministic window builder accounts for score pickup information,
-merges piano events and duplicate `(start, pitch)` notes, and chooses the first
-complete protocol-compatible 16-bar window. **99 pieces use the opening window;
-one uses the first subsequent compatible window** because the opening does not
-meet the protocol. `challenge_test/manifest.json` records the source SHA256,
-window index, pickup, selection reason, note counts, and MIDI parsing diagnostics.
-No new selection or window extraction was performed for this release.
-## Format
-Times use a sixteenth-note grid. A complete window ends at step 272, with steps
-0--15 reserved for an optional pickup. The four regular prompt bars occupy
-steps 16--79. Prompt notes have start in 0--79, and reference notes in 80--271.
-Pitch is an integer in 0--127; duration is a positive integer and notes do not
-extend beyond step 272. Notes are sorted by `(start, pitch)`.
-```json
-{"prompt": [{"start": 16, "pitch": 60, "duration": 4}]}
-```
-```json
-{"generation": [{"start": 80, "pitch": 64, "duration": 4}]}
-```
-The fixed-window MIDIs use 480 ticks per quarter note, 120 BPM, program 0 and
-velocity 64 for comparable listening. Original complete-piece assets retain
-their original contents. Genre labels and source previews are auxiliary;
-they are not additional generation conditions in the DAMN benchmark.
 ## Usage
 ```python
 from datasets import load_dataset
-benchmark = load_dataset("naivecynics/MetaScore-SFT-100", split="test")
-available = benchmark.filter(lambda row: row["content_available"])
-assert len(benchmark) == 100 and len(available) == 100
-```
-Pin the Hub revision for experiments. Use the same candidate count and sampling
-protocol when comparing results. DAMN's development evaluation generates eight
-candidates per prompt, averages their CLaMP3 cosine similarities to the reference
-within each prompt, and then averages across prompts; it does not select a
-reference-based best candidate. Confidence intervals resample prompts.
-## Provenance, permissions and limitations
-Metadata records 99 source scores as `cc-zero` and one as `publicdomain`.
-These are upstream declarations, not an independent verification of rights to
-every underlying composition or arrangement. Existing score credits and copyright
-notices are retained, and the explicit conflicting entry is included only in this private repository, without clearing its rights. There is
-no blanket license replacing source rights; see [LICENSE.md](LICENSE.md).
-The benchmark contains familiar repertoire and a small, curated selection of
-genres. Source metadata can be imperfect. References are one possible
-continuation rather than the only musically valid answer. Similarity to a
-reference is a development proxy, not a substitute for listening evaluation.
-The DAMN pipeline excludes benchmark matches from its training data; this does
-not establish the absence of overlap with other models' training corpora.
-## Citation and links
-Please credit both the original MetaScore work and this derived benchmark.
-The DAMN manuscript is being prepared; no arXiv identifier is assigned here.
 ```bibtex
 @inproceedings{xu2025metascore,
@@ -143,6 +88,7 @@ The DAMN manuscript is being prepared; no arXiv identifier is assigned here.
   booktitle={Proceedings of the International Society for Music Information Retrieval Conference},
   year={2025}
 }
 @misc{huang2026metascoresft100,
   title={MetaScore-SFT-100: A Fixed Piano Continuation Development Benchmark},
   author={Huang, Langchu and Xu, Yang},
@@ -152,11 +98,6 @@ The DAMN manuscript is being prepared; no arXiv identifier is assigned here.
 }
 ```
-- [MetaScore project and source distribution](https://github.com/wx83/MetaScore_Official)
-- [MetaScore paper](https://arxiv.org/abs/2410.02084)
-- [DAMN code](https://github.com/naivecynics/MIREX2026-DAMN/tree/mirex2026-submission)
-- [DAMN weights](https://huggingface.co/naivecynics/MIREX2026-DAMN)
-- [Official MIREX 2026 results](https://music-ir.org/mirex/wiki/2026:Symbolic_Music_Generation_Results)
-- [Official listening demo](https://futuremirex.com/demos/symbolic-music-generation-2026/)
-Maintainer: Langchu Huang (`langchuhuang@gmail.com`).

 # MetaScore-SFT-100
+## Overview
+**MetaScore-SFT-100** is a fixed, 100-piece benchmark for symbolic piano continuation, derived from [MetaScore](https://github.com/wx83/MetaScore_Official) and developed to evaluate [DAMN](https://github.com/naivecynics/MIREX2026-DAMN).
+Each example pairs a **4-bar piano prompt** (with an optional pickup) with a **12-bar reference continuation**, following the MIREX 2026 Symbolic Music Generation format. This is an independently curated development benchmark, **not the official MIREX test set**.
+## Dataset
+The benchmark contains **100 complete examples across 15 genre labels**, with fixed prompts, references, and source assets.
+| Resource | Contents |
 | --- | --- |
+| [`test.jsonl`](test.jsonl) | 100 entries with CID, genre, source, and prompt/reference note sequences |
+| [`challenge_test/`](challenge_test) | Fixed prompt/reference JSON, companion MIDIs, and window manifest |
+| `midi/`, `mscz/`, `mp3/` | Complete-piece MIDI, MuseScore scores, and existing source previews |
+| [`label/`](label) | Genre annotations |
+| `manifest.csv`, `summary.json` | Source selection metadata and statistics |
+| [`attribution.json`](attribution.json) | Per-piece attribution and source license records |
+| [`checksums.sha256`](checksums.sha256) | File integrity checksums |
+Notes use integer `(start, pitch, duration)` attributes on a sixteenth-note grid. Steps `0-15` are reserved for an optional pickup, prompt starts are in `0-79`, and reference starts are in `80-271`. Durations are positive and notes end no later than step `272`.
+Source permissions apply per piece; see [LICENSE.md](LICENSE.md). This repository is private, and the existing source-license conflict remains documented in [rights_conflicts.json](rights_conflicts.json).
+## Construction
+We selected piano scores from a genre-stratified sample of MetaScore using these criteria:
+- A single piano part, 4/4 meter, at least 16 measures, and no lyrics.
 - The upstream best-unique-arrangement filter.
+- A user rating of at least 4.5 from at least three ratings.
+Automatic filtering retained 106 candidates; manual listening removed six clear quality-control failures, leaving 100 pieces. Genre counts are not balanced.
+Deterministic extraction selects the first complete, protocol-compatible 16-bar window: **99 pieces use the opening window, and one uses a later window**. Fixed-window MIDIs use 480 ticks per quarter note, 120 BPM, and a fixed piano program. Complete-piece source assets retain their original contents; their MP3s are not renderings of the extracted windows.
+See [RECONSTRUCTION.md](RECONSTRUCTION.md) for reproduction instructions and [DETAILS.md](DETAILS.md) for full selection, timing, and provenance details.
 ## Usage
 ```python
 from datasets import load_dataset
+# Authenticate with an account granted access to this private repository.
+benchmark = load_dataset("naivecynics/MetaScore-SFT-100", split="test", token=True)
+assert len(benchmark) == 100
+example = benchmark[0]
+prompt = {"prompt": example["prompt"]}
+reference = {"generation": example["reference"]}
+```
+For reproducible experiments, pin the dataset revision and keep generation settings fixed across systems.
+**DAMN evaluation protocol:** generate eight continuations per prompt, compute each continuation's CLaMP3 cosine similarity to its reference, average across the eight candidates, then average across the 100 prompts. The 95% confidence intervals use bootstrap resampling over prompts. This is mean-of-eight, not best-of-eight. Reference similarity is a development proxy, not a substitute for listening evaluation.
+## Citation
+Please cite the original MetaScore work and this benchmark:
 ```bibtex
 @inproceedings{xu2025metascore,
   booktitle={Proceedings of the International Society for Music Information Retrieval Conference},
   year={2025}
 }
 @misc{huang2026metascoresft100,
   title={MetaScore-SFT-100: A Fixed Piano Continuation Development Benchmark},
   author={Huang, Langchu and Xu, Yang},
 }
 ```
+**Related resources:** [DAMN code](https://github.com/naivecynics/MIREX2026-DAMN) · [Model weights](https://huggingface.co/naivecynics/MIREX2026-DAMN) · [MIREX 2026 results](https://music-ir.org/mirex/wiki/2026:Symbolic_Music_Generation_Results) · [Listening demo](https://futuremirex.com/demos/symbolic-music-generation-2026/)
+Maintainer: Langchu Huang ([langchuhuang@gmail.com](mailto:langchuhuang@gmail.com)).

checksums.sha256 CHANGED Viewed

@@ -1,5 +1,6 @@
 fb17c2919d30257eb69f0399f9950eef45eeefb5076069a02043dc5e542008b5  LICENSE.md
-c22622d779ffedcbd712168e6f95f90dbaecc5ce0bc7531fe562061dee2fb00f  README.md
 3d7fb3d8ea12c6f947ad493b8ad7acd0b409d1ffeb8f0edfa6e6bd633b579760  RECONSTRUCTION.md
 343ab1ca9de0ad1388e326aba00574963099eb56419a1f5ed169f1cc530efe9e  attribution.json
 dd0b867a5f593be3d463c7e0414209f7cd8bd68ca3d93099b165585a18cde5d5  challenge_test/manifest.json

+9dacb0089cb22df76f57bebc55bc127efc95633312e973ab08cc1f2c7e6a4824  DETAILS.md
 fb17c2919d30257eb69f0399f9950eef45eeefb5076069a02043dc5e542008b5  LICENSE.md
+5b5f6c5c195841f017a93fb2c02d9212a3d7d553a7ad55b62a3a62cc2e1cb355  README.md
 3d7fb3d8ea12c6f947ad493b8ad7acd0b409d1ffeb8f0edfa6e6bd633b579760  RECONSTRUCTION.md
 343ab1ca9de0ad1388e326aba00574963099eb56419a1f5ed169f1cc530efe9e  attribution.json
 dd0b867a5f593be3d463c7e0414209f7cd8bd68ca3d93099b165585a18cde5d5  challenge_test/manifest.json