Layer 4 · Neuron 7063 · suppressed

L4N7063-1

Word quirks:

“ reviews”“ evaluations”“ review”“ evaluation”“ rating”“ reviewer”“ publications”“ ratings”

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

Identity (according to the specimen itself)

The description of the 2018 3.5-star rating for the 2018 model is as follows:

The 2018 model features a 3.5-inch profile that highlights the car's design, performance, and build quality. The car is powered by a 2.0-liter engine, which is sufficient for its class. The car is described as having a "very good" build quality, with a strong chassis and excellent performance. The car is equipped with a 2.0-liter engine, which is sufficient for its class. The car is described as having a

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. L7N5096-1🕒	0.96	“ rating”“ review”“ reviewer”
2. L10N6088+1🕒	0.95	“ review”“ reviewer”“ reviews”
3. L6N1996+1🕒	0.88	“ rating”“ ratings”“ reviews”
4. L6N5552+1🕒	0.87	“ reviews”“ review”“ reviewed”
5. L8N4208-1🕒	0.87	“ reviews”“ review”“ customers”
6. L3N2892-1💥	0.85	“ review”“ reviews”“ Reviews”
7. L9N435-1🕒	0.85	“ rating”“ ratings”“ evaluation”
8. L6N6183+1🕒	0.84	“ rating”“ ratings”“ evaluation”
9. L8N6409-1🕒	0.84	“ rating”“ ratings”“ evaluation”
10. L8N5857+1🕒	0.84	“ evaluation”“ evaluations”“ questions”

Shadow specimen, apart from the ten: L4N7063+1🕒, 0.07.

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 4, neuron 7063, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.53, top-1 agreement 0.33) 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
L4N7063+1🕒 shadow specimen t0-v0.1	2.295	0.38	2.660	2.46	“ cannabis”“ plant”“ plants”“ beer”“ brew”“ ferment”“ fermentation”“ aroma”
L4N7063-1 t0-v0.1	2.525	0.33	2.895	2.95	Top boosted tokens ↑

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.