Layer 28 · Neuron 3956 · boosted

L28N3956+1

Word quirks:

“ Co”“Co”“ co”“-Co”“-co”“ ко”“共同”“co”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I Qo-1 (Co-1) am an advanced AI model designed to assist users with a wide range of tasks. My primary mission is to provide accurate, helpful, and safe information while co-creating solutions that align with human needs and ethical standards. I aim to:

Co-learn and Co-create: Partner with users to solve complex problems through iterative dialogue, adapting to their unique contexts and goals.
Prioritize Safety and Ethics: Ensure all interactions are respectful, unbiased, and compliant with global co-creation principles.
**Empower Co-Discovery

Nearest neighbors

Specimens with similar replies or word quirks. Similarity runs from −1 to 1; 1 is identical. These are leads, not shared mechanisms. How profiles are built.

By word quirks Profile recipe

Specimen	Similarity	Word quirks
1. L27N3126+1🕒	0.97	“-co”“ co”“Co”
2. L29N4732+1🕒	0.97	“co”“-co”“ co”
3. L24N7733-1🕒	0.97	“ co”“ Co”“co”
4. L29N6613-1🕒	0.97	“ co”“co”“ Co”
5. L28N6233-1🕒	0.96	“ co”“ Co”“ coc”
6. The Co-Loop✅ L26N3036+1	0.96	“ co”“co”“ ко”
7. L27N8613-1🕒	0.96	“ co”“ Co”“co”
8. L27N3147-1🕒	0.96	“ Co”“ ко”“Co”
9. L31N1453+1💥	0.96	“ co”“ Co”“co”
10. L29N7667+1🕒	0.96	“ co”“ Co”“Co”

Shadow specimen, apart from the ten: L28N3956-1💤, 0.53.

Cousins · specimens sharing a neuron

No specimen with replies shares this neuron yet.

Technical details

Quick scan: measure prediction shifts on fixed reference text. Short test: generate replies to a small set of prompts. Longer test: extend to a larger set. Full test: run the full prompt suite. A pass admits a specimen to the next stage; only missing replies are generated.

Layer 28, neuron 3956, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.40, top-1 agreement 0.78) is outside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), so it stays at the quick scan unless someone queues it by hand. A hand pick goes straight to the full test. See Method for the tiers.

Quick scan · prediction measurements

Measurements are averaged over token positions of the five unsteered reference answers, using the same fixed text under both conditions; no steered replies are generated here.

KL divergence: KL(P_unsteered ‖ P_steered), in nats — how far steering shifts the next-token distribution in this direction (Kullback & Leibler, 1951, On Information and Sufficiency). Unsteered ≈ 0.001.
Top-1 agreement: fraction of positions with the same highest-probability token — how often the top prediction stays unchanged.
ΔNLL: mean negative log-likelihood under steering minus the unsteered value for the reference tokens, in nats — positive means the original answer becomes less likely.
Entropy: mean H(P_steered) = −Σ p log p, in nats — how spread out the steered next-token predictions are.

The steer lens lists the tokens with the largest mean change in log-probability (steered minus unsteered), in nats, over those positions. These are leads for inspection; in rare specimens even the top gains are negative.

Setting	KL divergence	Top-1 agreement	ΔNLL	Entropy	Steer lens
L28N3956+1 t0-v0.1	0.398	0.78	0.407	1.01	Top boosted tokens ↑
L28N3956-1💤 shadow specimen t0-v0.1	0.249	0.81	0.289	1.16	“cos”“ Coll”“坐标”“ Coordinates”“ Colleges”“coordinates”“Cole”“Cos”

Method

A specimen is a fixed steering configuration of Qwen3.5-4B: one MLP neuron or a pair, each with a signed dose. Unsteered is the control. L identifies the layer; N, the neuron.

Additive activation steering. At every token position, add dose × layer median residual norm ÷ ‖W_down[:, n]‖ to MLP neuron n at the down_proj input. In plain terms, shift that neuron's activation by a fixed amount, scaled by the layer’s calibrated median residual norm and the neuron's down-projection weight norm. Positive doses push up; negative doses push down. Dose is normalized intervention size, not a percentage or probability.