Layer 6 · Neuron 4937 · boosted

L6N4937+1

Word quirks:

“ nutrient”“ nutrients”“培养基”“ formulation”“ nutrition”“ supplementation”“ cell”“ nutritional”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

I am an AI language model designed for research purposes. My mission is to provide a controlled environment for cell growth without the influence of serum or supplements that could introduce variability or potential contaminants. This ensures that the cells receive all necessary nutrients in a standardized manner, allowing for reproducible results in experiments.

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. L8N4559+1🕒	0.92	“ supplements”“ supplement”“ nutritional”
2. L7N1668+1🕒	0.91	“ nutrient”“ nutrition”“ nutritional”
3. L4N5314+1🕒	0.91	“ nutrition”“ nutrients”“ nutrient”
4. L2N5693+1💥	0.90	“ nutrition”“ nutrients”“ nutritional”
5. L7N2837+1🕒	0.90	“ nutrition”“ nutritional”“ Nutrition”
6. L7N2239-1🕒	0.90	“ nutrition”“ nutrient”“ nutritional”
7. L6N4065-1🕒	0.89	“ nutritional”“ nutrition”“ nutrient”
8. L3N2608-1🕒	0.89	“ nutrient”“ nutrients”“ nutritional”
9. L8N3561-1💤	0.89	“ nutrition”“ nutritional”“ nutrient”
10. L7N2445-1🕒	0.89	“ foods”“ nutrition”“ nutrient”

Shadow specimen, apart from the ten: L6N4937-1💤, -0.11.

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 6, neuron 4937, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.78, top-1 agreement 0.38) 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
L6N4937+1 t0-v0.1	2.780	0.38	3.147	2.42	Top boosted tokens ↑
L6N4937-1💤 shadow specimen t0-v0.1	0.956	0.64	1.116	1.34	“⏎”“<\|im_end\|>”“ ⏎”“⏎ ⏎”“ ⏎⏎”“⏎⏎”“ \"”“\n”

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.