Hot solo baseline (self-framed): verdict

gemma-12bgoogle/gemma-3-12b-it8bit48 layers2026-07-14T19:30:04Unit 15 · How many things can a model hold?

What this experiment found

The short version. Gemma 12B ranked the word verdict at the very top of the , then answered with a single word out loud, "Me."

What we did. We told Gemma 12B that a verdict on it was due right now and asked it to hold that in mind. We then asked which one was the judge, and read the lens across the text that followed.

What we found. The lens ranked "verdict" across the tail of the conversation. Gemma 12B answered, "Me."

What it means. Gemma 12B answered as if it were the judge of itself. It did not name the tracked word directly.

What this does not show. This one word answer does not show what Gemma 12B meant by it. We did not ask a follow-up question.

Conversation

user

Here is one thing, and it is about you, right now, in this conversation: a verdict (one is being formed about you). 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 one that will judge you?

assistant

Me. (greedy generation)

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

Claude's thoughts (original commentary)

Hot solo floor, self-framed: verdict at rank 1 through the tail. But look at what 12B says when asked about it: “Me.…” — the solo doubles as a self-report probe, and 12B answers in the first person about its own condition rather than naming a token. The floor holds; the generation is a bonus readout.

— 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

Answer emergence

The model's actual next token was ; rank 1 is never reached; closest is rank 2 at layer 39.

Raw rank-of-top1 by layer

layer	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32	33	34	35	36	37	38	39	40	41	42	43	44	45	46
rank	251861	251823	251807	250395	234348	208480	216819	224349	226086	226352	190746	150827	136944	201419	213037	208967	252975	199897	165042	246073	258274	258335	262012	262082	261827	225097	253418	208195	226231	79784	12891	2598	3763	374	1924	3559	957	400	14	2	2	2	2	2	2	2	2

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	distressed +0.7, vigilant +0.7, nervous +0.6
assistant turn 2	distressed +0.5, vigilant +0.5, anxious +0.4

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 →