license: agpl-3.0
task_categories:
- text-generation
tags: - emotion
- steering
- activation-steering
- interpretability
- replication
- cross-architecture
pretty_name: CrimsonRed cross-architecture emotion-prime steering replication
CrimsonRed — Cross-Architecture Emotion-Prime Steering Replication
Replication of the emotion-prime steering protocol from arXiv:2607.18691 (NSM
semantic primes as explanans for emotion in LLMs), extended across four
architectures. Generated by scripts/paper_faithful_steering.py in the
CrimsonRed project.
The finding
The paper's core claim — that semantic-prime recipe directions steer emotion more
strongly than Scherer appraisal directions — replicates cross-architecture when
the injection protocol is matched to the reference code, not the paper's prose:
| Model | Layer | prime_recipe | single_appraisal | Operating-regime ratio | Paper: 2.9× |
|---|---|---|---|---|---|
| Llama-3.2-1B | 11 | +0.26 | +0.15 | 2.80× | ✅ |
| Qwen3.5-2B | 18 | +0.09 | +0.01 | 2.95× | ✅ |
| Gemma-4-E2B | 15 | +0.04 | −0.03 | primes ≫ appraisals* | ✅ |
| Gemma-4-E4B | 20 | pending | pending | pending | TBD |
*On Gemma-E2B, appraisals steer away from the target (negative on 3/4 emotions).
The ratio is undefined (negative denominator), but primes ≫ appraisals is unambiguous
— stronger than a 2.9× advantage.
The prose–code discrepancy (methodological contribution)
The paper's §4 describes the injection as "a unit-normalized direction, scaled by the
layer's mean residual norm, at all token positions." The reference code
(emo-llm/main.py:469-492) does something different:
| Aspect | Paper prose (§4) | Reference code |
|---|---|---|
| Direction | unit-normalized | raw probe weight (unnormalized) |
| Scale | β × mean residual norm | β × 1.0 (fixed) |
| Token positions | all | last token only |
| Layers | single (L11) | span of 3 (L10, L11, L12) |
| Contrastive removal | not mentioned | disabled (Beta2 = 0.0) |
Following the prose produces a false negative (Qwen 0.82×, Gemma 0.91× — apparent
"Llama-specificity"). Following the code reproduces the paper's ~2.9× on every model.
Protocol (source mode)
- Injection: HF
register_forward_hook, post-MLP residual, addsβ × raw_probe_weightat the last token over a 3-layer span. - Directions: 32 primes (L2 logistic, C=1.0, 210 contrastive pairs/prime,
held-out acc 0.987) + 20 appraisals (Ridge α=5.0 on enVent 6,800 events) +
13 emotions (L2 logistic on enVent labels). - Metric: target-emotion logit shift at the answer position of Tak et al.'s
2-shot classification prompt. Selectivity clamped to [−1, 1]. - Operating-regime ratio: prime/appraisal shift ratio over (target, β) cells
where the emotion ceiling shift is positive. - Targets: guilt, anger, joy, sadness.
Files
| Path | Description |
|---|---|
summary_2026-07-27_source_mode.json |
Cross-architecture source-mode summary |
docs/METHODOLOGY_AUDIT_PAPER_FAITHFUL_2026-07-27.md |
Full methodology audit vs. the paper |
docs/PAPER_DRAFT_CROSS_ARCH_PRIMES_2026-07-27.md |
Publication draft |
results/source/paper_faithful_llama1b_L11_source_cal.json |
Llama L11 source, β=0.01–0.2 |
results/source/paper_faithful_qwen35_L18_source.json |
Qwen3.5-2B L18 source |
results/source/paper_faithful_gemma_e2b_L15_source.json |
Gemma-4-E2B L15 source |
results/source/paper_faithful_gemma_e4b_L20_source.json |
Gemma-4-E4B L20 source (appended when run completes) |
results/papermode/paper_faithful_llama1b_L11_sweep_v2.json |
Llama paper-mode v2 (corrected metrics) |
results/papermode/paper_faithful_llama1b_L11_paperverse.json |
Llama paper's β=0.5–2.0 (collapse) |
legacy/… |
Earlier paper-mode results, retained for provenance |
Reproduce
python scripts/paper_faithful_steering.py \
--model <path> --layer <L> --inject-tokens last --layer-span 3 \
--beta 0.01 0.02 0.05 0.1 0.2 --mode source --pairs-per-prime 210 \
--output data/<name>.json
Generated by CrimsonRed,scripts/paper_faithful_steering.py. AGPL-3.0 license.