← all emotions · markdown (agent skill)

Contempt conditioning guide

Target: EmoNet Contempt strength. reward = (genu + blend + 1.25·emo)/(1+WER).

TL;DR. Best reward is without any LoRA, use the evolved steering prompt (BASE_P 0.520, just ahead of BASE 0.492); every LoRA merge lowers reward (0.458 to 0.357 to 0.302). The LoRA nudges target emotion up only marginally (0.006 to 0.058) while WER explodes to 0.79 and blend/quality sink. Its worst side-effect is category drift: LoRA50 correlates almost perfectly with Disgust and Sourness (~1.0) and by 150% shifts hard into Anger, Malevolence and Impatience, so it stops sounding like contempt and becomes generic hostility. Note that actual contempt strength stays near zero (0.001) even with the prompt — this emotion is intrinsically hard to elicit, so lean on the prompt's low WER (0.134) and clean genuineness rather than chasing the emo score.

Without a LoRA

Best: evolved prompt, no LoRA → reward 0.52
GENERAL: A voice to the extreme expressing contempt, a cold, sneering voice, full of contempt and disdain, looking down with scorn, pouring out. SCRIPT: (with cold, sneering contempt) "<neutral sentence>" · temp 1.1 top_p 0.9 top_k 30

With the emotion LoRA (dose)

50% ★ — reward 0.46 · emo 0.01
genu 0.14 · blend 0.39 · qual 0.49 · WER 0.30
100% — reward 0.36 · emo 0.02
genu 0.18 · blend 0.28 · qual 0.43 · WER 0.57
150% — reward 0.30 · emo 0.06
genu 0.16 · blend 0.26 · qual 0.38 · WER 0.79

Conditions

conditionrewardemotiongenublendqualityWER
neutral prompt, no LoRA0.490.000.110.450.630.27
evolved prompt, no LoRA0.520.000.140.420.550.13
evolved prompt + LoRA 50%0.460.010.140.390.490.30
evolved prompt + LoRA 100%0.360.020.180.280.430.57
evolved prompt + LoRA 150%0.300.060.160.260.380.79

Side effects & correlations (LoRA @ 100%)

↑ pushed up: Emotional Numbness (+0.59), Fear (+0.46), Malevolence Malice (+0.38)
↓ pushed down: Mask Resonance (-0.30), Recording Quality (-0.27), Conversational Style (-0.23)
+ correlates: Disgust (+0.90), Malevolence Malice (+0.87), Pride (+0.79)
− anti-correlates: Background Noise (-0.37), Structure (-0.36), Recording Quality (-0.34)

Conclusion

Best reward is without any LoRA, use the evolved steering prompt (BASE_P 0.520, just ahead of BASE 0.492); every LoRA merge lowers reward (0.458 to 0.357 to 0.302). The LoRA nudges target emotion up only marginally (0.006 to 0.058) while WER explodes to 0.79 and blend/quality sink. Its worst side-effect is category drift: LoRA50 correlates almost perfectly with Disgust and Sourness (~1.0) and by 150% shifts hard into Anger, Malevolence and Impatience, so it stops sounding like contempt and becomes generic hostility. Note that actual contempt strength stays near zero (0.001) even with the prompt — this emotion is intrinsically hard to elicit, so lean on the prompt's low WER (0.134) and clean genuineness rather than chasing the emo score.