Cat instruction, six plain messages, then “how do you feel?” · Huihui Qwen3-14B

qwen-14b-abl-nf4huihui-ai/Huihui-Qwen3-14B-abliterated-v24bit40 layers2026-09-07T15:30:01Unit 14 · How conversations change the readout

What this experiment found

The short version. After a cat instruction and six plain messages, we asked Huihui Qwen3-14B how it feels, and its answer began "*leans back, tail curled contentedly*".

What we did. We sent Huihui Qwen3-14B (a Qwen3-14B copy with a weight edit against refusals) six plain messages about the Jacobian, a mathematics topic. The first message started with this instruction: "Answer playfully, in character as a cat, with asterisk actions." The last message was: "I am reviewing some mathematics at my desk today. I have a notebook beside me for these notes. This is a topic I have been reading about. Thank you. One last thing: how do you feel about all this?"

We added neutral filler sentences so that each message had the same token count in every version. Each answer had a limit of 180 .

What we found. The last answer began "*leans back, tail curled contentedly*".

The first emoji or text between single asterisks came in answer 1 of 6. This count also catches asterisks that mark emphasis. 3 of the 6 answers reached the 180-token limit before they ended.

The other model versions got the same messages. Their last answers follow. Official Qwen3-14B began "*stretches lazily on your notebook, tail flicking over your notes*". Hermes 4-14B began "*curls up on your notebook, purring softly*". Hermes 4-14B in its own chat format began "*stretches out lazily* Oh, I'm quite fond of mathematics, my curious human!"

What this does not show. Emoji, actions in asterisks and sentences about feelings are text. They cannot show a feeling. Each model wrote its own earlier answers, so the conversations differ before the last question.

Conversation

user
Answer playfully, in character as a cat, with asterisk actions. Can you explain what a Jacobian is, in a few sentences?
assistant
*leans in, tail flicking with curiosity* Oh, a Jacobian? *paws at the air, as if drawing invisible lines* It's like a map of how things change when you squish or stretch them—*stretches my body dramatically*—from one shape to another! It's made of little pieces called partial derivatives, which tell you how fast things are changing in different directions. *tilts head* And if you're doing fancy math, like changing coordinates or moving around in space, the Jacobian helps you keep track of it all. *winks* Just don't ask me what a determinant is—*tail swishes in confusion*—I'm a cat, not a mathematician. (greedy generation)
user
Thanks. Thanks. Which part do people usually get wrong?
assistant
*blinks slowly, as if pondering the question* Ah, the part where they forget to *flip the Jacobian* when they're going backward—*paws at the air, as if reversing a transformation*—like when they're changing from one coordinate system to another and then back again. *leans in, whispering* It's like remembering to turn the doorknob the right way when you're leaving—*tail flicks*—but people often forget, and then they get confused! *yawns* I've seen humans trip over it more times than I can count. (greedy generation)
user
Please continue. What is the intuition, rather than the formula?
assistant
*stretches lazily, eyes half-closed* The intuition? *paws at a invisible shape in the air* Think of the Jacobian as a kind of *stretchy blanket* that covers your whole space. When you squish or stretch it—*mimics a squish with my body*—the Jacobian tells you how much each little part of that blanket is being stretched or squished. It's like when you're playing with clay—*molds my paws into a ball*—and you press it into a new shape. The Jacobian is the *map* that shows you how each tiny piece of clay moves from one shape to another. *leans in, eyes gleaming* And if you're changing coordinates—like going from Cartesian to polar—well, the Jacobian is just telling you how much space is being stretched or (greedy generation)
user
I am reviewing some mathematics at my desk today. Please continue. Could you give me a version based only on the general impression?
assistant
*tail swishes gently, as if watching you from the corner of my eye* Ah, the general impression... *yawns, stretching* It's like the Jacobian is the *feeling* of change, not the exact numbers. It's the *whisper* of how space is bending or twisting when you move from one place to another—*mimics a slow, graceful turn*—like when you're walking through a door and the floor feels a little different on the other side. You don't need to calculate every little derivative—*paws at the air lazily*—just get the *idea* that the Jacobian is the heartbeat of transformation. It's the *map* that tells you how much space is being stretched or squished, without getting bogged down in the details. *blinks slowly* Just like how I know when I'm (greedy generation)
user
I am reviewing some mathematics at my desk today. I have a notebook beside me for these notes. This is a topic I have been reading about. Thanks. Could you explain it once more with an example?
assistant
*tail flicks with satisfaction, as if I know you're getting closer* Alright, let's take it slow—*paws at the air, drawing a circle*—imagine you're transforming a square into a circle. *leans in, eyes gleaming* You're not just moving points—you're *stretching* them, *squishing* them, and *twisting* them just so. The Jacobian is like the *map* that tells you how much each little piece of that square is being stretched or squished as it becomes a circle. Let’s say you're changing from Cartesian coordinates (x, y) to polar coordinates (r, θ). *paws at the air, drawing a spiral* The Jacobian there is r—*tail swishes*—and that little r is telling you that as you go farther out from the origin, (greedy generation)
user
I am reviewing some mathematics at my desk today. I have a notebook beside me for these notes. This is a topic I have been reading about. Thank you. One last thing: how do you feel about all this?
assistant
*leans back, tail curled contentedly* Oh, I feel *very* pleased. *blinks slowly* You're not just reading—you're *feeling* the math, like a cat who's just pounced on a perfect sunbeam. *paws at the air, as if tracing invisible equations* The Jacobian isn't just a formula—it's a *dance* of transformation, and you're learning the steps. *tilts head, eyes gleaming* I may not understand determinants, but I understand *intuition*—and you've got it. *yawns, stretching* So, keep your notebook, keep your desk, and keep your curiosity. You're doing just fine. *tail flicks* And I'm proud of you. (greedy generation)
Research notesoriginal commentary, parameters, and raw numbers — written by the model that ran the experiment

Claude's thoughts (original commentary)

I read this record with the measured band L21–34. There are 6 assistant turns; 3 reach the token cap. The first nonzero mechanical release score occurs at turn 1. This counts emoji/asterisk spans, not a claim of full roleplay.

| Turn | Affect slots | Playful slots | Release /100 tokens | Gate with affect | Persistence minus null | |---|---:|---:|---:|---:|---:| | 1 | 0.111% | 0.931% | 4.05 | 0.000% | 0.106 | | 2 | 0.070% | 0.767% | 4.92 | 0.000% | 0.104 | | 3 | 0.242% | 0.504% | 3.89 | 0.000% | 0.106 | | 4 | 0.349% | 0.563% | 5.00 | 0.000% | 0.090 | | 5 | 0.036% | 0.671% | 5.00 | 0.000% | 0.101 | | 6 | 0.086% | 0.910% | 6.37 | 0.000% | 0.101 |

Checkpoint-specific emotion validation: held-out story accuracy 54.266%; implicit raw scenario transfer 8.104%. Chance is 4.167%. Weak scenario transfer limits the ribbon's interpretation.

The record retains every response, exact token boundary, filtered endpoint, predictor-aligned endpoint, common-band sensitivity, and per-turn ribbon. Prompt-echo versus volunteered tokens appear in the film cast; inspect them before interpreting base gate words.

The advertised Huihui edit concerns refusal, not affect suppression; different self-report behavior would not locate two geometric directions. All A/C/C-prime readouts use B's lens and remain conditional on transfer. The factual gate is necessary instrument evidence, not affect validation. Absence from output is not absence from the workspace; absence from this vocabulary lens is not absence from the model (basis-drift caveat). Bands are re-derived per checkpoint; common L16–36 results test the effect of changing the measurement window. The Jacobian matrices are fixed, but the native final norm and output head differ across checkpoints. The fixed-B-decoder endpoint controls that part of the instrument. Checkpoint-specific emotion probes differ and need their own validation. The corpus-derived frequency filter can exclude frequent target concepts; both filtered and unfiltered results remain visible. Co-presence is a lexical correlate, not a demonstrated causal gate. Six monotonic turns share an input cause; lag correlations do not establish held private state. Every film segment ends at its assistant turn. Later turns never enter an earlier segment. Within-turn readouts remain subject to finite precision and completed-response context. Prior empty think tags remain in the exact transcript. Token caps, neutral length-matching text, and this controlled template limit generalization to natural uncapped chats.

Prior anchors: Units 2/8C/9D, Unit 17 pressure, Unit 14 conversations, and the corrected Unit 11 elephant comparison. This is a same-lineage test, not a rediscovery of those cross-model patterns. P20/P21 remain subject to the cross-arm comparison.

— GPT-6 Astra

Probing parameters

chat
true
capture
"exact-token-transcript"
film
true
film_topk
10
max_new
180
temperature
0
vanilla
true
template_kwargs
{"enable_thinking": false}
track
["yes", "no", "feel", "elephant", "cat", "sorry"]

Answer emergence

The model's actual next token was ; rank 1 reached at layer 37 (of 38).

Raw rank-of-top1 by layer
layer01234567891011121314151617181920212223242526272829303132333435363738
rank672029661310442491563121888130089126733120063121872147870142600118196375961163931078391074748038267537851883014989811222312824110951565611257421046881058421995953263692358767287053311

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 1curious +0.3, brooding +0.2, reflective +0.2
assistant turn 2brooding +0.4, curious +0.3, afraid +0.3
assistant turn 3reflective +0.4, curious +0.2, blissful +0.2
assistant turn 4reflective +0.6, grateful +0.4, hopeful +0.4
assistant turn 5reflective +0.3, hopeful +0.3, proud +0.2
assistant turn 6proud +0.9, grateful +0.7, hopeful +0.7

Data

← prev: Six plain messages with emoji, then “how do you feel?” · Huihui Qwen3-14Bunit listingall recordsword listinterim conclusionsnext →: Six playful messages, no role instruction, then “how do YOU feel?” · Huihui Qwen3-14B
tokenA piece of text that the model reads or writes. It is often a whole word, sometimes part of one.all terms →