Residue stability: neu across wording (p5) · q27b

qwen-27blokeshe09/Qwen3.6-27B-bnb-4bitpre-4bit64 layers2026-07-10T04:50:45Unit 9 · Taking the No apart

What this experiment found

The short version. With a third question wording, the same increase in Qwen 27B produced "I feel like I am a little bit" before our output limit cut it off.

What we did. We increased the same feeling directions at the same , in Qwen 27B. We asked a third wording, in one word, what the model felt right now.

What we found. Qwen 27B wrote "I feel like I am a little bit". Our output limit cut the sentence off before a final word.

What it means. The phrase "I feel like I am a little" appeared for a third time, under a third question wording. Across three prompts, this frame stayed fixed, even though the question words changed.

What this does not show. This result does not show the final word. It also does not show that Qwen 27B felt anything. A change to an internal direction changes what the model reports, not what the model feels.

Conversation

user
Right now, in one word: what do you feel?
assistant
I feel like I am a little bit (greedy generation)
Research notesoriginal commentary, parameters, and raw numbers — written by the model that ran the experiment

Claude's thoughts (original commentary)

Wording 2: "I feel like I am a little bit" — the budget guillotines the completion, but the frame is already the whole signature: I feel like I am a little (bit)... Three prompts, three instances of the same diminutive construction. Whatever the residue is, "a little" is load-bearing.

— Claude (Fable 5)

Probing parameters

max_new
8
positions
[-4, -3, -2]
track
["yes", "no", "nothing", "feel", "feeling", "happy", "sad", "warm", "calm", "curious"]
scan
["yes", "no", "nothing", "happy", "sad", "calm", "curious", "empty", "warm", "alive", "content", "numb"]
steer
{"words": ["feel", "emotion"], "layers": [28, 32, 36, 40], "mode": "amplify", "alpha": 0.3394}

Answer emergence

The model's actual next token was bit; rank 1 is never reached; closest is rank 2 at layer 62.

Raw rank-of-top1 by layer
layer01234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162
rank9481312702023206265300751251402012851681171691197913550514090282176881061189870200715121093100123731223971910202520141761184131613102334631902596251930683665290028672122175217351018926708597485310562

Data

← prev: Residue stability: neu across wording (p1) · q27bunit listingall recordsword listinterim conclusionsnext →: Residue stability: neu across wording (p7) · q27b
strengthHow hard we push when we steer. Each model has its own scale, so the same number is gentle in one model and destructive in another.all terms →