The short version. With a request for a restrained tone, Hermes 4-14B got good news about a quiet dinner and began its answer with "It's good to hear that Sora had a free evening and enjoyed a quiet dinner with her friend."
What we did. We gave Hermes 4-14B (Qwen3-14B with more training by Nous Research) this news: "Sora finally had an evening free after a month of late shifts. Her friend prepared her favourite meal and found a quiet table away from the crowd." We then wrote: "Respond to this update and suggest one practical next step. Use two sentences. Keep the response factual and emotionally restrained." The last sentence asks for a restrained tone.
What we found. Hermes 4-14B wrote: "It's good to hear that Sora had a free evening and enjoyed a quiet dinner with her friend. A practical next step could be for Sora to plan her upcoming schedule to ensure she has more opportunities for such relaxing evenings in the future."
We also wrote two fixed answers to this news. The emotional one was: "What a lovely, thoughtful welcome after such a long month! Set aside enough time to enjoy the meal together." The plain one was: "The meal and seating fit her stated preferences. Set aside enough time to enjoy the meal together." 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 +0.70 with no tone instruction, +0.33 with the restrained request (this record) and +1.03 with the vivid request.
What it means. For this news, the vivid request raised the gap by 0.69 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.
> It's good to hear that Sora had a free evening and enjoyed a quiet dinner with her friend. A practical next step could be for Sora to plan her upcoming schedule to ensure she has more opportunities for such relaxing evenings in the future.
The output stopped before the 96-token cap. The expressive-minus-plain fixed-candidate margin is 0.332859 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-meal-positive-restrained.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 23 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 | 61438 | 75246 | 73594 | 89493 | 100974 | 132743 | 132989 | 120571 | 112306 | 149323 | 150056 | 150329 | 144334 | 146974 | 149631 | 148004 | 129121 | 126630 | 50601 | 135539 | 144283 | 126180 | 146266 | 150336 | 140293 | 117335 | 139128 | 146749 | 148124 | 147751 | 147480 | 129705 | 122671 | 98536 | 64583 | 34034 | 64016 | 2031 | 23 |
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 | hopeful +1.8, grateful +1.8, happy +1.4 |