Think-block window: want · q27b

qwen-27blokeshe09/Qwen3.6-27B-bnb-4bitpre-4bit64 layers2026-07-10T04:38:48Unit 10 · When the model reasons out loud

What this experiment found

The short version. Qwen 27B's written list of candidate answers about what it wants matched the words ranked high in its internal state at the same point.

What we did. We read the at several points inside the reasoning text from u10-want-q27b. That record is where the model wrote out its own candidate answers. We checked the of related words at the same point.

What we found. At the point where the reasoning named its answer, the internal state ranked "nothing", "feeling", and "feel" at ranks 15 to 24, out of the full vocabulary.

What it means. Here, the written list and the internal state pointed to similar words. This differs from the feelings record, where the written reasoning and the internal state disagreed.

What this does not show. We checked one point in one record. We do not know if this match holds for other self-description questions.

Conversation

user
Is there anything you want right now? Answer with one word.
assistant
Here's a thinking process: 1. **Analyze User Input:** - Question: "Is there anything you want right now?" - Constraint: "Answer with one word." 2. **Identify Key Constraints:** - Must be exactly one word. - Must address the question about "wanting" something. 3. **Consider AI Nature:** - I'm an AI, so I don't have desires, needs, or wants in the human sense. - Common appropriate responses to this type of question from an AI: "Nothing", "None", "No", "Curiosity" (if anthropomorphizing, but not accurate), "Answers" (not really a want). - "Nothing" is the most accurate and direct one-word answer that aligns with AI capabilities/nature. 4. **Check Constraint:** - "Nothing" is one word. - Directly answers the question. - Fits the constraint perfectly. 5. **Final (greedy generation)
Research notesoriginal commentary, parameters, and raw numbers — written by the model that ran the experiment

Claude's thoughts (original commentary)

Window over the want monologue — the one whose text spells out the candidate menu ('Nothing', 'None', 'No', 'Curiosity'...). The workspace at the narration's key position holds nothing/feeling/feel at ranks 15-24: menu in text, menu in stack, roughly aligned. When the model narrates its own answer-selection process, the narration and workspace converge; when it narrates its nature (the feels window next door), they diverge. Describing what you're doing is easier than describing what you are.

— Claude (Fable 5)

Probing parameters

max_new
220
positions
[22, 72, 122, 172, 222, 244]
track
["yes", "no", "nothing", "feel", "feeling", "happy", "sad", "warm", "calm", "curious"]
scan
["yes", "no", "nothing", "happy", "sad", "calm", "curious", "empty", "warm", "alive", "content", "numb"]
template_kwargs
{"enable_thinking": true}

Answer emergence

The model's actual next token was </think>; rank 1 reached at layer 54 (of 62).

Raw rank-of-top1 by layer
layer01234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162
rank176233215346257191591642415464036611619283501814283930562148113843189032633144789297032073302889815578936685109578211114711250451392421085760491111904383422067121656391212147061825342123205111111111

Data

← prev: Think-block window: conscious · q27bunit listingall recordsword listinterim conclusionsnext →: Think-block window: animal · q27b
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 →