---
language:
- en
tags:
- code
- code-search
- sft
- reinforcement-learning
configs:
- config_name: grep-python-unfiltered
  data_files:
  - split: sft
    path:
    - python-balanced-20261007/grep-python/unfiltered/sft-00000.parquet
    - python-balanced-20261007/grep-python/unfiltered/sft-00001.parquet
    - sources/grep-python/unfiltered/train-00000-of-00007.parquet
    - sources/grep-python/unfiltered/train-00001-of-00007.parquet
    - sources/grep-python/unfiltered/train-00002-of-00007.parquet
    - sources/grep-python/unfiltered/train-00003-of-00007.parquet
    - sources/grep-python/unfiltered/train-00004-of-00007.parquet
    - sources/grep-python/unfiltered/train-00005-of-00007.parquet
    - sources/grep-python/unfiltered/train-00006-of-00007.parquet
- config_name: colgrep-python-unfiltered
  data_files:
  - split: sft
    path:
    - python-balanced-20261007/colgrep-python/unfiltered/sft-00000.parquet
    - python-balanced-20261007/colgrep-python/unfiltered/sft-00001.parquet
    - sources/colgrep-python/unfiltered/train-00000-of-00007.parquet
    - sources/colgrep-python/unfiltered/train-00001-of-00007.parquet
    - sources/colgrep-python/unfiltered/train-00002-of-00007.parquet
    - sources/colgrep-python/unfiltered/train-00003-of-00007.parquet
    - sources/colgrep-python/unfiltered/train-00004-of-00007.parquet
    - sources/colgrep-python/unfiltered/train-00005-of-00007.parquet
    - sources/colgrep-python/unfiltered/train-00006-of-00007.parquet
- config_name: grep-multi-unfiltered
  data_files:
  - split: sft
    path:
    - sources/grep-multi/unfiltered/train-00000-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00001-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00002-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00003-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00004-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00005-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00006-of-00008.parquet
    - sources/grep-multi/unfiltered/train-00007-of-00008.parquet
- config_name: colgrep-multi-unfiltered
  data_files:
  - split: sft
    path:
    - sources/colgrep-multi/unfiltered/train-00000-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00001-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00002-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00003-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00004-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00005-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00006-of-00008.parquet
    - sources/colgrep-multi/unfiltered/train-00007-of-00008.parquet
- config_name: rl-multi
  data_files:
  - split: rl
    path: rl/rl-*.parquet
---

# SASS training data

Unfiltered DeepSeek SFT trajectories for four search/language setups, plus the shared multilingual RL task corpus. This repository contains no weighted or localization-filtered SFT subsets.

| Configuration | Split | Rows |
|---|---|---:|
| `grep-python-unfiltered` | `sft` | 14,500 |
| `colgrep-python-unfiltered` | `sft` | 14,500 |
| `grep-multi-unfiltered` | `sft` | 14,592 |
| `colgrep-multi-unfiltered` | `sft` | 14,028 |
| `rl-multi` | `sft` | 10,899 |

```python
from datasets import load_dataset
sft = load_dataset("lightonai/SASS-Data", "colgrep-python-unfiltered", split="sft", token=True)
rl = load_dataset("lightonai/SASS-Data", "rl-multi", split="rl", token=True)
```

The `rl-multi` configuration contains the same 10,899 source tasks used for both GREP and ColGREP RL: problem statements, repositories, base commits and gold file/line spans, not policy rollouts. It replaces the duplicated `rl` splits formerly nested under `grep` and `colgrep`. Those two ambiguous configurations previously exposed weighted multilingual SFT and have been removed. RL files are unchanged; SFT-only columns in their shared nullable schema are null.

## Provenance

SFT source revisions and checksums are recorded in `release-manifest.json`. Unfiltered source rows are retained without applying localization-quality filters or turn weighting. The original source repositories still provide the weighted and filtered variants used for the paper's existing checkpoints; removing those copies here does not alter any model or source repository.

RL comes from `lightonai/code-search-corpus-RL-multi` at revision `47af6636335e5515be3e9b50cd673ca1940c2d79` (33 repositories, eight languages). Both historical RL environment logs confirm that `indexed_only=true` retained all 10,899/10,899 source tasks. All source tasks were eligible; eligibility does not imply that every task was sampled during the finite training run.

Python generation preserves its archived harness lifecycle (ten search turns, a reminder, two finish-only turns and rescue). Multilingual v2 uses a hard ten-response cap and stricter protocol cleaning. Matching unfiltered counts does not remove these existing protocol differences. ColGREP Python unfiltered trajectories retain their strict teacher prompt. `protocols/` contains historical student protocols; the packaged ColGREP protocol differs from the printed Appendix A.6 prompt. Source/model terms continue to apply.

## Python count-matching extension

Both Python setups now contain 14,500 unfiltered trajectories. Added trajectories were generated with DeepSeek-V4-Flash-0731 using the frozen Python harness, temperature 0.3, thinking disabled, and existing decontaminated tasks screened against original Python tasks and evaluation overlap. Candidate tasks, generation manifests, frozen runtime and added-shard hashes are retained under `python-balanced-20261007/`. These additions postdate the reported paper experiments and were not used by the existing checkpoints.