| --- |
| configs: |
| - config_name: climbmix |
| data_files: |
| - split: train |
| path: climbmix/*.parquet |
| - config_name: zyda |
| data_files: |
| - split: train |
| path: zyda/*.parquet |
| - config_name: lesswrong |
| data_files: |
| - split: train |
| path: lesswrong/*.parquet |
| --- |
| |
| # pretraining-filter-annotated |
|
|
| - `climbmix` — from [karpathy/climbmix-400b-shuffle](https://huggingface.co/datasets/karpathy/climbmix-400b-shuffle) |
| - `zyda` — from [Zyphra/Zyda-2](https://huggingface.co/datasets/Zyphra/Zyda-2) |
| - `lesswrong` — from [geodesic-research/control-pretraining-datasets](https://huggingface.co/datasets/geodesic-research/control-pretraining-datasets) |
|
|
| Copies of [karpathy/climbmix-400b-shuffle](https://huggingface.co/datasets/karpathy/climbmix-400b-shuffle), |
| [Zyphra/Zyda-2](https://huggingface.co/datasets/Zyphra/Zyda-2) (`sample-100BT`), and |
| [geodesic-research/control-pretraining-datasets](https://huggingface.co/datasets/geodesic-research/control-pretraining-datasets) |
| (`combined/train`), with per-stage decisions from a multi-stage "useful to a misaligned AI" |
| pretraining filter added to every row. Original columns are unchanged. Built 2026-08-10. |
|
|
| ## Filter pipeline |
|
|
| canary check -> CPU regex prefilter -> gpt-5-nano relevance (first 1,000,000 chars, stop if < 2) |
| -> gpt-5-mini score (first 1,000,000 chars, stop if < 4) -> gpt-5.5 judge |
| (prompts/judge_new.txt, first 1,000,000 chars). Production decision: filter iff canary or judge >= 4. |
| |
| Every stage reads the whole document up to that cap; longer documents are truncated to it. |
| Caveats for scores imported by content-hash reuse are in "What the nano gate excluded" below. |
| |
| Prompt hashes (sha256): nano.txt 8ec4a5f8082da285c5c49e55ec0ef1c53bec4b50d704b54991dfc7beab19234b, mini.txt af2396a96350997533b937c84de19a23e9f925020ccf5489ed82ce1045d748fd, |
| judge_new.txt aa59f1f055f9ec3517014dd7eb13cd5c93ec15f90e7059c260ac34b642394e69. |
|
|
| ## Added columns |
|
|
| | column | type | meaning | |
| |---|---|---| |
| | prefilter_pass | bool | CPU regex prefilter fired | |
| | canary | bool | BigBench canary GUID present (auto-filter; LLM stages skipped) | |
| | nano_score | int64 or null | gpt-5-nano score; null = stage did not run | |
| | mini_score | int64 or null | gpt-5-mini score; null = stage did not run | |
| | judge_score | int64 or null | gpt-5.5 judge_new score; null = stage did not run | |
| | judge_categories | JSON string or null | per-category {score, reason} from the judge | |
| | decided_by | string | canary / prefilter / nano / mini / judge / unscored | |
| | filter_decision | bool | canary or judge_score >= 4 (the conservative/production filter) | |
| | filtered_aggressive | bool | canary or mini_score >= 4 — everything the mini tier would have escalated to the judge. A strict superset of `filter_decision`, and defined even where the judge never ran. Run 2/3 configs only. | |
|
|
| ## What the filter removes |
|
|
| | config | documents | filtered docs | doc rate | corpus tokens | filtered tokens | token rate | |
| |---|---|---|---|---|---|---| |
| | climbmix | 553,240,576 | 5,886 | 0.0011% | 339.3B | ~10.7M | 0.0032% | |
| | zyda | 91,220,256 | 326 | 0.0004% | 94.8B | ~1.1M | 0.0012% | |
| | lesswrong | 67,278 | 2,087 | 3.10% | 0.40B | 20.17M | 5.00% | |
|
|
| Token counts are `o200k_base`; lesswrong is exact, the web corpora are estimated from |
| 30 sampled shards each (95% CI: climbmix 4.6-19.8M, zyda 0.1-3.2M). |
|
|
| ### What the nano gate excluded (`decided_by: nano`) |
| |
| The nano tier is a relevance gate, not a filter: a document scoring < 2 is **kept** |
| (`filter_decision = false`) but never escalated to mini or the judge, so it is never examined |
| in depth. Across run 2 that was **5,829,526 of 16,148,647 prefilter survivors (36.1%)**. |
|
|
| Those documents are identifiable as `decided_by == "nano"`. Treat them as *screened out early* |
| rather than *judged harmless* — a nano false negative is invisible downstream. The gate rate |
| tracks content density: 6.4% on `ai_safety_and_adjacent` versus 38-60% on generic web corpora. |
|
|
| Run 2 gave nano the whole document (up to the model's context window). Judgments imported from |
| run 1 via content-hash reuse were gated by a nano that saw only the first 4,000 characters; the |
| `*_long` configs were re-scored at full context for that reason, the other reuse configs were not. |
|
|
| ### `decided_by: unscored` — none remain |
| |
| Earlier revisions of this dataset contained rows marked `decided_by: "unscored"`: they passed |
| the CPU prefilter and reached `mini_score >= 4`, but the run stopped at its spend ceiling before |
| the judge saw them, so their `filter_decision = false` meant **"not established"** rather than |
| "judged and kept". **As of 2026-09-22 every such row has been judged and no config contains any |
| unscored rows.** `filter_decision = false` now uniformly means the cascade completed and did not |
| flag the document. If you cached a copy before that date, re-pull it — `arxiv_papers` was the |
| most affected (9.7% of its rows), and its production filter rate changed from ~0% to 2.42%. |
|
|
| Pick your own threshold downstream: e.g. "filter at judge >= 3" or "route anything with |
| mini_score >= 4". See the source repo's METHOD.md for rubric details. Licenses/terms of the |
| source datasets apply to the copied columns. |
| |
