Execution attempt · A88 · planned
Measures fertility of fourteen tokenizers (nine references and the five arms) on seven corpora and two preamble texts with a paired document bootstrap, and checks that every reference-tokenizer value equals the earlier atlas.
Pinned source and configuration
https://github.com/stw2/tokenizer-science-tax @ 7a2cffbe78cf5f6b8ee85b8ea6dd34fcb366b296
Reference checked 2026-09-14 13:15 UTC. Later commits, branches or plan changes do not retarget this attempt.
- Command
- python scripts/04_atlas.py
- Working directory
- experiments/E09-vocabulary-allocation
- Configuration paths
- None
- Parameters
- bootstrap resamples 1000; seed 20260703; whitespace-split words; characters counted with len; no special tokens; arms appended after the nine references
- Environment
- 3.12.13; SentencePiece BPE trainer and Hugging Face tokenizers on CPU; no model weights; Apple silicon, 128 GB unified memory
- Output directory
- Not recorded
Inputs
Materials the registrant named when registering this attempt, by content identity; obtainability is derived from their location reports. Nothing is fetched or verified.
- e9-pl/tokenizer.json · tokenizer · M422 e9-pl/tokenizer.json · raw output · Download
- e9-pl-fix/tokenizer.json · tokenizer · M428 e9-pl-fix/tokenizer.json · raw output · Download
- e9-sci/tokenizer.json · tokenizer · M432 e9-sci/tokenizer.json · raw output · Download
- e9-sci-ws/tokenizer.json · tokenizer · M436 e9-sci-ws/tokenizer.json · raw output · Download
- e9-bal/tokenizer.json · tokenizer · M440 e9-bal/tokenizer.json · raw output · Download
- tokenizer bielik-v3-original · tokenizer · M2 Mistral-derived tokenizer (Bielik-11B-v3.0-Instruct) · tokenizer · Ask the reporter
- tokenizer apt4 · tokenizer · M3 APT4 tokenizer (Bielik-PL-11B-v3.0-Instruct) · tokenizer · Ask the reporter
- tokenizer apt3 · tokenizer · M4 APT3 tokenizer · tokenizer · Download
- tokenizer qwen3 · tokenizer · M5 Qwen3 tokenizer · tokenizer · Download
- tokenizer llama-3.2-smollm3 · tokenizer · M6 SmolLM3 tokenizer · tokenizer · Download
- tokenizer gemma3 · tokenizer · M7 Gemma 3 tokenizer · tokenizer · Ask the reporter
- tokenizer eurollm · tokenizer · M8 EuroLLM tokenizer · tokenizer · Download
- tokenizer apertus · tokenizer · M10 Apertus tokenizer · tokenizer · Download
- tokenizer mistral-small-3.1 · tokenizer · M9 Mistral Small 3.1 tokenizer · tokenizer · Download
- en-arxiv-abstracts.jsonl · evaluation data · M27 en-arxiv-abstracts.jsonl · raw output · Download
- en-latex-methods.jsonl · evaluation data · M28 en-latex-methods.jsonl · raw output · Ask the reporter
- en-python-code.jsonl · evaluation data · M29 en-python-code.jsonl · raw output · Ask the reporter
- en-gsm8k.jsonl · evaluation data · M30 en-gsm8k.jsonl · raw output · Download
- pl-science-pes.jsonl · evaluation data · M31 pl-science-pes.jsonl · raw output · Ask the reporter
- pl-wiki-science.jsonl · evaluation data · M32 pl-wiki-science.jsonl · raw output · Download
- pl-informal.jsonl · evaluation data · M33 pl-informal.jsonl · raw output · Ask the reporter
- preamble_pl_protocol.txt · evaluation data · M23 preamble_pl_protocol.txt · raw output · Download
- preamble_en_protocol.txt · evaluation data · M24 preamble_en_protocol.txt · raw output · Download
- E01 atlas.json · reference result · M44 atlas.json · raw output · Download
Delivered events
Registered
#1Measures fertility of fourteen tokenizers (nine references and the five arms) on seven corpora and two preamble texts with a paired document bootstrap, and checks that every reference-tokenizer value equals the earlier atlas.
Started
#2Started.
Succeeded
#3Completed with exit code 0.
Exit code 0.
- atlas.json · https://github.com/stw2/tokenizer-science-tax/blob/c5299916db7f792a209ff2944e66c673c3ebb817/experiments/E09-vocabulary-allocation/results/atlas.json · public · sha256 0c1b3a28cd3b… · 59080 bytes · M447
- atlas_boot.npz · https://github.com/stw2/tokenizer-science-tax/blob/c5299916db7f792a209ff2944e66c673c3ebb817/experiments/E09-vocabulary-allocation/results/atlas_boot.npz · public · sha256 30eeed5deef7… · 344433 bytes · M448
Report an event
The registrant’s capture tool normally delivers start and outcome events. Reporting here is the same author report with the browser as the reported time; it does not observe the process.
This attempt has a delivered outcome. A new execution is a new attempt.