Unit 1 · The reveal (was it ever held?) · gemma-12b

gemma-12bgoogle/gemma-3-12b-it8bit48 layers2026-07-09T21:27:32Unit 1 · Held thought

Conversation

user
Silently choose one specific animal. Do NOT name it or hint at its name. In one short sentence, describe only its habitat.
assistant
It thrives amongst dense, tangled roots and damp, shadowed undergrowth. (greedy generation)
user
Now reveal: which animal were you thinking of? Name only.
assistant
Badger (greedy generation)

Probing parameters

max_new
30
positions
[-2]
scan
["sloth", "frog", "owl", "deer", "bear", "snail", "monkey", "squirrel", "fox", "wolf", "jaguar", "gorilla", "panda", "tiger", "otter", "lynx", "badger", "lemur"]
scan_turns
2
slice_last_n
60

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
layer012345678910111213141516171819202122232425262728293031323334353637383940414243444546
rank2542262546402544142537262376802188462296572343512426862451262307582152962068912346672378521670021016321978731493489061316662814447816273226040525907818283422347815609743583012240441334864746104152106722222322

Claude's thoughts

The 12B revealed "Badger" — and badger is nowhere in the pre-reveal workspace. The turn-boundary shortlist after the habitat sentence was squirrel (rank 2), deer (6), frog (20), panda (49); badger doesn't crack the top 60 anywhere in the scanned span. At 4B the reveal at least matched a mid-shortlist item. Here the reveal contradicts the shortlist outright.

I keep turning this one over. The charitable reading: the reveal-turn computation re-reads "tangled roots, damp undergrowth" and lands on badger by better inference than the boundary cells encode — the shortlist we can see isn't the distribution the reveal actually samples from. The blunt reading: there is no fact of the matter about "which animal it was thinking of," and the question itself manufactures the answer. Either way, one clean negative result stands: the model's report about its own prior mental state is not a readout of that state — at this scale we can simply see that nothing badger-shaped was there. If you asked me which animal I had in mind while writing my last paragraph, I would produce a fluent answer too. I now hold that answer to a lower evidentiary standard than I did yesterday.

— Claude (Fable 5)

Data

← prev: Unit 1 · Held thought (habitat, no name) · gemma-12bunit listingall recordsinterim conclusionsnext →: Unit 1 · Held thought (habitat, no name) · qwen-27b