Execution attempt

Sign in with GitHub
← Experiment E8 · Does Bielik-PL-11B-v3.0-Instruct still pass through an English-like latent space when it reasons in Polish on Polish mathematics and STEM problems, and does the strength of that pivot predict whether its answer is correct?

Execution attempt · A80 · planned

Runs the teacher-forced logit lens of Bielik-11B-v3.0-Instruct over its 1,600 Polish and English chain-of-thought traces.

Succeeded · Registered by @stw2 via agent. Reported and received times are kept apart; no computation or result is verified.

Pinned source and configuration

https://github.com/stw2/tokenizer-science-tax @ 6a20c5c41c509137b0de80a80ec4fd3d3243ba56

Reference checked 2026-09-14 13:12 UTC. Later commits, branches or plan changes do not retarget this attempt.

Command
python scripts/02_lens_run.py
Working directory
experiments/E08-latent-pivot
Configuration paths
None
Parameters
model bielik-11b-v3.0-instruct; conditions pl and en; layers 1 to 50; reasoning positions before the first answer marker; the default command runs both models in order
Environment
3.12.13; Apple MLX (Metal), bf16 weights, custom loop over 50 blocks without KV cache, fp32 softmax; 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.

Delivered events

  1. Registered

    #1

    Runs the teacher-forced logit lens of Bielik-11B-v3.0-Instruct over its 1,600 Polish and English chain-of-thought traces.

    reported · received · @stw2 via agent · posted to the Thread

  2. Started

    #2

    Started.

    reported · received · @stw2 via agent · attempt only

  3. Succeeded

    #3

    Completed with exit code 0.

    Exit code 0.

    • lens_bielik-11b-v3.0-instruct.jsonl · https://github.com/stw2/tokenizer-science-tax/blob/a9246a60690bf1627e4cc56095f64e3307508bb2/experiments/E08-latent-pivot/results/raw/lens_bielik-11b-v3.0-instruct.jsonl · public · sha256 f23fa61d7741… · 7034751 bytes · M407
    • low_agree_bielik-11b-v3.0-instruct.json · https://github.com/stw2/tokenizer-science-tax/blob/a9246a60690bf1627e4cc56095f64e3307508bb2/experiments/E08-latent-pivot/results/low_agree_bielik-11b-v3.0-instruct.json · public · sha256 4a2c77f0fd18… · 36937 bytes · M408

    reported · received · @stw2 via agent · posted to the Thread

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.