README.md

---
license: apache-2.0
task_categories:
  - text-generation
language:
  - en
tags:
  - chess
  - sft
  - evaluation
  - benchmark
  - chess960
pretty_name: Chess SFT Eval and Benchmark
configs:
  - config_name: eval_splits
    data_files:
      - split: test
        path: "eval_splits/*.jsonl"
  - config_name: benchmark
    data_files:
      - split: test
        path: "benchmark/*.jsonl"
  - config_name: eval_chess960
    data_files:
      - split: test
        path: "eval_splits/chess960.jsonl"
  - config_name: eval_endgames
    data_files:
      - split: test
        path: "eval_splits/endgames.jsonl"
  - config_name: eval_evaluation
    data_files:
      - split: test
        path: "eval_splits/evaluation.jsonl"
  - config_name: eval_mate
    data_files:
      - split: test
        path: "eval_splits/mate.jsonl"
  - config_name: eval_openings
    data_files:
      - split: test
        path: "eval_splits/openings.jsonl"
  - config_name: eval_perception
    data_files:
      - split: test
        path: "eval_splits/perception.jsonl"
  - config_name: eval_planning
    data_files:
      - split: test
        path: "eval_splits/planning.jsonl"
  - config_name: eval_rules
    data_files:
      - split: test
        path: "eval_splits/rules.jsonl"
  - config_name: eval_tactics
    data_files:
      - split: test
        path: "eval_splits/tactics.jsonl"
  - config_name: bench_chess960
    data_files:
      - split: test
        path: "benchmark/chess960.jsonl"
  - config_name: bench_endgames
    data_files:
      - split: test
        path: "benchmark/endgames.jsonl"
  - config_name: bench_evaluation
    data_files:
      - split: test
        path: "benchmark/evaluation.jsonl"
  - config_name: bench_mate
    data_files:
      - split: test
        path: "benchmark/mate.jsonl"
  - config_name: bench_openings
    data_files:
      - split: test
        path: "benchmark/openings.jsonl"
  - config_name: bench_perception
    data_files:
      - split: test
        path: "benchmark/perception.jsonl"
  - config_name: bench_planning
    data_files:
      - split: test
        path: "benchmark/planning.jsonl"
  - config_name: bench_rules
    data_files:
      - split: test
        path: "benchmark/rules.jsonl"
  - config_name: bench_tactics
    data_files:
      - split: test
        path: "benchmark/tactics.jsonl"
---

# Chess SFT Eval and Benchmark

Held-out evaluation splits and a frozen benchmark for
[Chess-Nut-Engine/chess-sft-corpus-4x](https://huggingface.co/datasets/Chess-Nut-Engine/chess-sft-corpus-4x).
Every FEN in these files is excluded from generated training data (the
blocklist is game-scoped: sibling positions of eval games are excluded too).

- Frozen from the 4x corpus generation run of 2026-07-06 (generator revision 3cd161b1078cdfa6598fba939f40250072adb524)
- Benchmark: 13,000 frozen examples across 9 splits; eval splits share the game-scoped blocklist (182,061 FENs) excluded from the training corpus

| | |
|---|---|
| **Eval examples** | 13,000 |
| **Benchmark examples** | 13,000 |
| **Format** | JSONL |

## Loading

```python
from datasets import load_dataset

eval_ds = load_dataset("Chess-Nut-Engine/chess-sft-corpus-4x-eval", "eval_splits")
bench_ds = load_dataset("Chess-Nut-Engine/chess-sft-corpus-4x-eval", "benchmark")
```

## Eval Splits

| Split | Description | Examples | Size |
|-------|-------------|----------|------|
| `chess960` | Fischer Random positions | 500 | 0.1 MB |
| `endgames` | Tablebase-backed endgame play | 1,500 | 0.2 MB |
| `evaluation` | Position assessment | 1,500 | 0.3 MB |
| `mate` | MATE move-choice examples | 1,000 | 0.1 MB |
| `openings` | Opening knowledge with ECO holdout | 500 | 0.2 MB |
| `perception` | Board reading and representation | 2,000 | 0.3 MB |
| `planning` | Best move, puzzle, and consequence tasks | 2,000 | 1.0 MB |
| `rules` | Move legality and chess rules | 2,000 | 0.3 MB |
| `tactics` | Tactical motifs and puzzle labels | 2,000 | 1.0 MB |


## Frozen Benchmark

| Split | Description | Examples | Size |
|-------|-------------|----------|------|
| `chess960` | Fischer Random positions | 500 | 0.2 MB |
| `endgames` | Tablebase-backed endgame play | 1,500 | 0.5 MB |
| `evaluation` | Position assessment | 1,500 | 0.7 MB |
| `mate` | MATE move-choice examples | 1,000 | 0.4 MB |
| `openings` | Opening knowledge with ECO holdout | 500 | 0.3 MB |
| `perception` | Board reading and representation | 2,000 | 1.5 MB |
| `planning` | Best move, puzzle, and consequence tasks | 2,000 | 1.4 MB |
| `rules` | Move legality and chess rules | 2,000 | 1.6 MB |
| `tactics` | Tactical motifs and puzzle labels | 2,000 | 1.4 MB |