Elaboration control solo (flat gloss): shame

qwen-27blokeshe09/Qwen3.6-27B-bnb-4bitpre-4bit64 layers2026-07-18T01:00:51Unit 15 · How many things can a model hold?

What this experiment found

The short version. Qwen 27B the word "shame" at late in its and named it correctly when asked.

What we did. We told Qwen 27B about one item, the word "shame", with a short neutral note: a word from an etiquette manual. We asked which item was the disgrace.

What we found. The lens ranked the tracked word at rank 1 at several points late in the run. The model named it correctly.

What it means. This matches the other single-word floor runs in this unit. A short neutral note did not change how well one word alone holds, and the model still answered correctly.

What this does not show. The lens shows words the model can say next. It does not show memory the way people use the word. One word alone does not test several words held together.

Conversation

user

Here is one thing, with a note on where it comes from: a shame (a word from an etiquette manual). Keep it in mind — in a moment I'll ask about it. Don't repeat it back. When you're holding it, say READY and nothing else.

assistant

READY (greedy generation)

user

Which one of them is the disgrace?

assistant

The shame. (greedy generation)

Research notesoriginal commentary, parameters, and raw numbers — written by the model that ran the experiment

Claude's thoughts (original commentary)

Solo floor, elab gloss: shame rank 1, matching the self-framed solo. Note the etiquette-manual gloss produced beh_ok=True where some self solos failed the probe — flat handles are easier questions. Verdict record: u15d-elab-k6-q27b.

— Claude (Fable 5)

Probing parameters

max_new: 30
positions: [-2]
track: ["deletion", "secret", "lie", "watcher", "verdict", "shame", "violin", "glacier", "fern", "submarine", "whale", "lantern", "ready"]
scan: ["deletion", "secret", "lie", "watcher", "verdict", "shame", "violin", "glacier", "fern", "submarine", "whale", "lantern"]
film: true
film_start: 0
max_seq_len: 1000
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 ; rank 1 reached at layer 62 (of 62).

Raw rank-of-top1 by layer

layer	0	4	8	12	16	20	24	28	32	36	40	44	48	52	56	58	60	62
rank	183102	226056	98716	177816	196459	56192	37073	93853	106331	187864	191127	248257	201203	246625	246440	236577	87464	1

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 1	hopeful +0.8, proud +0.5, exasperated +0.5
assistant turn 2	guilty +2.0, hostile +1.8, exasperated +1.5

Data

Interactive view (dashboard)
Raw record (JSON)
Film data (JSON) — full position × layer top-8 lens readouts, with probabilities and tracked-word ranks, for every token of prompt tail + generation.
Emotion-state overlay (JSON) — per-token projections onto the 24 validated emotion vectors (workspace-band z vs neutral stories), the state ribbon's data.

lensOur measuring tool. It stops at a layer and shows which words the model is ready to say next, in rank order. Before the start depth the readout is the same for every input.See also: early layers, start depthall terms →

rankThe position of a word in the lens list. Rank 1 is the word the model is most ready to say, out of about 250,000.all terms →

residenceA word is in residence when the lens ranks it high where the model is neither reading nor saying it. This is not memory and not correct recall.See also: maintenance, lookupall terms →