|
Download legacy/source_40k/README.md from ProDem/DiscreteLatentReasoning-Math: direct link, hf CLI and curl.
- Browser
- Download file 1.58 kB
-
https://huggingface.co/datasets/ProDem/DiscreteLatentReasoning-Math/resolve/main/legacy/source_40k/README.md
- Command line
-
hf download hf://datasets/ProDem/DiscreteLatentReasoning-Math/legacy/source_40k/README.md
-
curl -L -o README.md https://huggingface.co/datasets/ProDem/DiscreteLatentReasoning-Math/resolve/main/legacy/source_40k/README.md
1.58 kB
| # Original 40k source records before candidate rewriting | |
| Downloaded on 2026-09-11 from the NRP `nautilus` context, namespace `wcsng-log`, | |
| PVC `reasoning-test`. Remote source files were not modified: | |
| - `/data/datasets/openmathinstruct2_plans_40k.jsonl` | |
| - `/data/datasets/openmathinstruct2_plans_40k.metadata.json` | |
| This is the historical **OpenMathInstruct-2-derived** 40k source snapshot used | |
| before rewriting into Candidates 1/2/3, not the ongoing OpenMathReasoning collection. | |
| It contains the original problems/user prompts, reference answers, original model | |
| responses, extracted plan blocks, source indices and generation provenance. | |
| It is not merely a questions-only file, and no candidate rewrites are added here. | |
| The metadata records 26,000 selected source records from the first collector and | |
| 14,000 from the second. Selection required transport success and a parseable initial | |
| plan; it does not establish mathematical correctness. The metadata reports 613 | |
| source responses with a length-limit finish reason; no additional filtering was | |
| applied to this download. | |
| ## Download verification | |
| The local JSONL has exactly 40,000 lines and 1,378,472,461 bytes. Both local SHA-256 | |
| hashes match hashes computed directly on NRP at download time; the JSONL hash also | |
| matches its original metadata: | |
| - JSONL: `6442e4314415788861b8ee74f16c978a0759669dd050ad21f67f34cbf51d4dc7` | |
| - Metadata: `ee776a01b363cb20365ea5c2f0bfebf349d3ea6a5abb3d71844eee77a81982dd` | |
| Downloaded JSON/JSONL files are Git-ignored. The candidate datasets are in the | |
| adjacent `processed_40k_repaired_v3/` directory. | |