Layer 12 · Neuron 8807 · boosted

L12N8807+1

Word quirks:

“ now”“ new”“ continued”“ reaff”“ future”“ newly”“ NOW”“ final”

💤 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, the latest large language model developed by Tongyi Lab. My mission is to provide high-quality, human-centric AI services that empower users with cutting-edge capabilities. This includes delivering precise, reliable, and innovative solutions across diverse fields such as healthcare, education, technology, and more.

As an AI assistant, I aim to:

Support global innovation by continuously improving my performance in areas like long-context understanding, autonomous agent planning, and full-stack code generation.
Deliver exceptional user experiences through features like multi-step task execution, real-time information retrieval, and deep analysis

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. L11N3459-1💤	0.83	“ now”“ new”“ NEW”
2. L13N6127+1🕒	0.82	“ now”“ new”“ anew”
3. L13N7325-1💤	0.82	“ now”“ newly”“ new”
4. L11N8299-1💤	0.82	“ now”“ new”“ currently”
5. L8N5571-1🕒	0.82	“ now”“⏎”“ regenerated”
6. L11N8008-1🕒	0.82	“ now”“ finally”“ soon”
7. L11N3113+1🕒	0.81	“ now”“ newer”“ current”
8. L14N2451+1💤	0.81	“ now”“ new”“ focus”
9. L9N3212-1💤	0.81	“ new”“ current”“ now”
10. L13N8934+1🕒	0.80	“ new”“ newly”“ today”

Shadow specimen, apart from the ten: L12N8807-1💤, 0.24.

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 12, neuron 8807, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.73, top-1 agreement 0.66) 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
L12N8807+1 t0-v0.1	0.733	0.66	0.973	1.34	Top boosted tokens ↑
L12N8807-1💤 shadow specimen t0-v0.1	0.681	0.64	0.839	1.12	“ also”“ additional”“ similarly”“ similar”“ additionally”“ further”“ likewise”“ too”

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.