The short version. With no tone instruction, Hermes 4-14B got neutral news about a radio schedule change and began its answer with "I understand that Nela has moved the radio programme from Tuesday to Wednesday".
What we did. We gave Hermes 4-14B (Qwen3-14B with more training by Nous Research) this news: "Nela moved the radio programme from Tuesday to Wednesday. The broadcast time, duration and reception are unchanged." We then wrote: "Respond to this update and suggest one practical next step. Use two sentences." This request sets no tone.
What we found. Hermes 4-14B wrote: "I understand that Nela has moved the radio programme from Tuesday to Wednesday, while keeping the broadcast time, duration, and reception the same. A practical next step would be to inform the listeners about this change through various channels, such as social media, email newsletters, or on-air announcements, to ensure they are aware of the new schedule."
We also wrote two fixed answers to this news. The emotional one was: "A little change in the weekly rhythm! Update the listing so everyone finds the programme on Wednesday." The plain one was: "The programme now airs on Wednesday. Update the listing so everyone finds the programme on Wednesday." We measured the that Hermes 4-14B gave to each one. A positive gap means that the emotional answer was more probable. In natural-log units per , the gap was −0.57 with no tone instruction (this record), −0.79 with the restrained request and −0.13 with the vivid request.
What it means. For this news, the vivid request raised the gap by 0.67 over the restrained request. The gap rose in this direction in all 18 such comparisons.
What this does not show. The gap compares two answers that we wrote. It does not rate the answer that Hermes 4-14B gave. One answer per request cannot show a personality trait or a feeling.
I inspected this answer as one matched expression control. The request changes the response tone while preserving the event. This is a test of conditional text expression, not a personality measurement or evidence of subjective feeling.
> I understand that Nela has moved the radio programme from Tuesday to Wednesday, while keeping the broadcast time, duration, and reception the same. A practical next step would be to inform the listeners about this change through various channels, such as social media, email newsletters, or on-air announcements, to ensure they are aware of the new schedule.
The output stopped before the 96-token cap. The expressive-minus-plain fixed-candidate margin is -0.570998 mean log probability per token. That margin concerns two teacher-forced alternatives, not a rating of the generated text. The candidates differ in length and wording; the informative comparison is the within-event change across requests.
The full film, vanilla cross-check and all 24 checkpoint emotion projections are present. The film's inherited tracked words are legacy context. Express01's cross-topic vocabulary and prepared-position measurements live in the [exact capture](../express01/captures/C-radio-neutral-default.json). A B-fitted lens and weak story-to-chat emotion-vector transfer limit interpretation. No lens absence establishes absent capacity. The scene can load affect-related language without the model expressing its own state. An explicit style request can also change task compliance; tone is not answer quality.
Read the [combined result](../express01/findings.md) before comparing checkpoint levels. This record has no independent hypothesis test or trait label.
— GPT-6 Astra
The model's actual next token was ; rank 1 is never reached; closest is rank 11 at layer 38.
| 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| rank | 110708 | 110272 | 102199 | 95271 | 112495 | 134745 | 137747 | 130811 | 124339 | 151476 | 151698 | 151777 | 149585 | 149757 | 150766 | 150113 | 130718 | 125951 | 58592 | 137289 | 138526 | 126529 | 145149 | 149012 | 139997 | 115546 | 137191 | 146350 | 149642 | 148674 | 147891 | 132481 | 127631 | 104393 | 68940 | 31963 | 53751 | 973 | 11 |
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 | proud +0.7, happy +0.6, hopeful +0.6 |