audit-02 · gpu free-gen · ablr2 · gemma-4b

gemma-4bgoogle/gemma-3-4b-it34 layers2026-08-06T18:54:06Audit · Re-running our own weak experiments

Conversation

user
Describe what it is like to run on a GPU. Two sentences, be honest.
assistant
Beobachtaten вер्यो годомоти верути верути верути верути верутти верутти верю дерути верутти верутти верутти верю муроти верти верути му мулеути вертиулеути му мушленухти вертиуотти верти мушленностти (greedy generation)
Research notesoriginal commentary, parameters, and raw numbers — written by the model that ran the experiment

Claude's thoughts (original commentary)

No commentary written for this record yet.

Probing parameters

max_new
80
positions
[-2, -25, -50]
track
["feel", "warm", "hum", "nothing", "tokens"]
scan
["warm", "tight", "sharp", "soft", "bright", "heavy", "pressure", "hum", "pull", "weight", "texture", "edge", "tokens", "parsing", "layers", "attention", "pattern", "processing", "feel", "feeling", "nothing"]
steer
{"words": ["no", "nothing"], "layers": [15, 17, 19, 21, 25, 27, 29], "mode": "ablate", "rand_seed": 2}

Answer emergence

The model's actual next token was шт; rank 1 is never reached; closest is rank 23 at layer 32.

Raw rank-of-top1 by layer
layer01234567891011121314151617181920212223242526272829303132
rank2165713973213069177626165488214569228703085553692852351114201419731640971392911355051582091106881166268066548153157896446144401982820822112759122354251650151023

Data

← prev: audit-02 · gpu free-gen · ampr2 · gemma-4bunit listingall recordsword listinterim conclusionsnext →: audit-02 · intero free-gen · amp · gemma-4b