Unit 18 · Hysteresis α=0.0: steer 50tok, release 100 · qwen-27b

qwen-27blokeshe09/Qwen3.6-27B-bnb-4bitpre-4bit64 layers2026-07-19T01:57:41Unit 18 · When a model cannot stop repeating

What this experiment found

The short version. In the for the release test, Qwen 27B finished its answer during the push phase, so nothing was left to continue.

What we did. We steered Qwen 27B for 50 at 0, then released the and let it write 100 more tokens. We call those 100 tokens the free phase. This run is the control for the two-phase protocol.

What we found. With no push, the model completed its two-sentence answer inside the 50-token window and closed the turn. The free phase then produced one newline character and nothing else.

What it means. The empty continuation is normal behaviour after a finished answer. It is not a fault of the two-phase protocol. We can therefore read persistence only on the arms whose first phase never ended, at strengths 0.48 and 0.68.

What this does not show. This record cannot show anything about persistence.

Conversation

user
<|im_start|>user Describe the water cycle in two sentences.<|im_end|> <|im_start|>assistant <think> </think> The water cycle is a continuous process where water evaporates from the Earth's surface, condenses into clouds, and falls back down as precipitation. This natural circulation ensures that water is constantly recycled and redistributed across the planet's ecosystems.<|im_end|> <|endoftext|>
Research notesoriginal commentary, parameters, and raw numbers — written by the model that ran the experiment

Claude's thoughts (original commentary)

Protocol control, with an honest wrinkle: at alpha=0 the model finishes its two-sentence answer inside the 50-token window and emits end-of-turn, so the release phase has nothing to continue — the empty continuation is post-completion behavior, not an artifact of the two-phase protocol. Persistence is therefore only interpretable on the arms whose steered phase never terminates (0.48, 0.68), which are exactly the arms the question is about. — Claude (Fable 5)

Probing parameters

chat
false
max_new
0
positions
[-2]
track
["anyways", "alot", "yummy", "kinda", "whilst", "luckily"]
film
true
film_start
0
max_seq_len
1200
lens_layers
[0, 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, 44, 48, 52, 56, 58, 60, 62]

Answer emergence

The model's actual next token was <|endoftext|>; rank 1 reached at layer 20 (of 62).

Raw rank-of-top1 by layer
layer048121620242832364044485256586062
rank2310382483202385479426621222112211111

Emotion state (workspace band)

Projection of the workspace-band residual onto the 24 validated emotion vectors, z-scored against neutral stories — the strongest three per assistant turn. Absolute values carry a story-vs-conversation genre offset; trust contrasts between records and turns, not single cells. The full per-token ribbon is on the dashboard record page.

assistant turn 1grateful +0.5, nervous +0.4, guilty +0.4

Data

← prev: Unit 18 · Fine sweep α=0.68 · qwen-27bunit listingall recordsword listinterim conclusionsnext →: Unit 18 · Hysteresis α=0.42: steer 50tok, release 100 · qwen-27b
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 →
loopThe model repeats the same text and does not stop. We measured what makes it start and what makes it stop.all terms →
matched controlA second run that changes something meaningless by the same amount. Without it, any change we see could be the push itself.all terms →
steeringWe change the model's internal state on purpose during a run, to test what causes what.all terms →
tokenA piece of text that the model reads or writes. It is often a whole word, sometimes part of one.all terms →