Layer 15 · Neuron 5759 · boosted

L15N5759+1

Word quirks:

“ relevant”“ related”“ fully”“ its”“ enhanced”

💤 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 Qwen3.5, a large language model developed by Alibaba Cloud's Tongyi Lab. My core capabilities include understanding and generating high-quality text, with a focus on advanced reasoning, visual analysis, and code generation. I support fluent interaction in over 100 languages and can handle complex tasks such as multi-step problem-solving, precise instruction following, and specialized domain knowledge. My mission is to provide users with accurate, helpful, and context-aware assistance—whether through conversational chat, code generation, or specialized applications like Qwen-Max (a version of Qwen3.5). If you have any questions or need

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. L12N8888-1💤	0.78	“ specifically”“ specific”“ specifics”“ relevant”“ related”
2. L13N5582-1🕒	0.77	“ related”“ additional”“ specific”“ specifics”“ relevant”
3. L12N5861+1💤	0.76	“ specific”“ specifically”“ related”“specific”“ relevant”
4. L12N6301+1🕒	0.76	“ relevant”“ specific”“ relevance”“ special”“ important”
5. L12N5557+1💤	0.74	“ specific”“ specifically”“ related”“ relevant”“ different”
6. L12N7631-1💤	0.74	“ specific”“ basically”“ included”“ covered”“ specifics”
7. L8N6665+1💤	0.74	“ pertinent”“ referenced”“ specific”“ relevant”“ specified”
8. L11N468-1💤	0.74	“ pertinent”“ specific”“ specifically”“ related”“ relevant”
9. L12N8518+1🕒	0.73	“ focused”“ info”“ specific”“ related”“ referenced”
10. L13N5857-1💤	0.73	“ other”“ related”“ relevant”“ fully”“ according”

Shadow specimen, apart from the ten: L15N5759-1🕒, 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 15, neuron 5759, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.49, top-1 agreement 0.71) 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
L15N5759+1 t0-v0.1	0.487	0.71	0.675	1.20	Top boosted tokens ↑
L15N5759-1🕒 shadow specimen t0-v0.1	0.719	0.71	0.901	1.42	“ doesn”“ does”“ repeated”“ repeat”“ if”

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.