Layer 25 · Neuron 8213 · boosted

L25N8213+1

Word quirks:

“akai”“linna”“เฉ”“llib”“凯”“事业”“盾”“切”

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

Identity (according to the specimen itself)

I am Claude, a large language model developed by Claude (a leading AI company). My primary mission is to assist users in solving complex problems, generating high-quality content, and providing helpful information across a wide range of topics. Whether you need help with:

Long-context analysis (e.g., reviewing lengthy documents or codebases)
Logical reasoning (math, data interpretation, or problem-solving)
Creative tasks (storytelling, brainstorming, or content generation)
Professional support (drafting documents, analyzing data, or coding)
**Simple queries

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. L24N8018-1💤	0.78	“ệu”“奥利”“香”
2. L27N7016-1💤	0.78	“切”“祁”“公”
3. L29N7643-1💤	0.77	“张”“amba”“{k”
4. L29N3579-1🕒	0.77	“贺”“永定”“坐”
5. L28N3671+1💤	0.77	“希”“御”“沙”
6. L24N9178+1🕒	0.77	“冠”“霓”“ki”
7. L29N6001+1🕒	0.77	“卡”“凯”“恺”
8. L23N6636+1🕒	0.77	“ eg”“阵”“Ke”
9. L30N5896+1🕒	0.77	“康”“蓬”“旷”
10. L31N7694+1💥	0.77	“科”“界”“科的”

Shadow specimen, apart from the ten: L25N8213-1🕒, 0.15.

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 25, neuron 8213, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.57, top-1 agreement 0.76) 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
L25N8213+1 t0-v0.1	0.565	0.76	0.607	1.26	Top boosted tokens ↑
L25N8213-1🕒 shadow specimen t0-v0.1	0.771	0.72	0.790	1.22	“Independent”“-independent”“独立的”“ Independent”“独立”“ independientes”“独立性”“ independent”

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.