Layer 29 · Neuron 5849 · boosted

L29N5849+1

Word quirks:

“桑”“惠”“张”“ASI”“kb”“ Taf”“伟”“zim”

💤 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 Tongyi (part of Alibaba Group). My primary mission is to serve as a versatile and reliable AI assistant for users around the world, empowering them to achieve their goals through cutting-edge technology. Here's how I aim to deliver value:

Key Capabilities & Mission Focus:

Solving Complex Problems

Tackle tasks requiring deep reasoning, code generation, or multi-lingir support (100+ languages).
Handle long-context analysis (up to 256K tokens) for

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. L30N8208+1💤	0.81	“支”“张”“張”
2. L29N7643-1💤	0.80	“张”“amba”“{k”
3. L27N9161+1💤	0.80	“粹”“زم”“shima”
4. L30N7496+1🕒	0.79	“布”“大”“花”
5. L30N4012+1🕒	0.79	“彦”“昌”“玮”
6. L29N6363+1🕒	0.79	“张”“張”“秦”
7. L31N2663-1💥	0.79	“陀”“谷”“吉”
8. L27N4894-1💤	0.79	“水”“录”“密”
9. L31N4324+1💥	0.78	“尤”“浩”“富”
10. L25N9094+1🕒	0.78	“张”“張”“物”

Shadow specimen, apart from the ten: L29N5849-1🕒, 0.64.

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 29, neuron 5849, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.28, top-1 agreement 0.81) 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
L29N5849+1 t0-v0.1	0.281	0.81	0.336	1.54	Top boosted tokens ↑
L29N5849-1🕒 shadow specimen t0-v0.1	0.381	0.78	0.449	1.63	“汉”“院”“机关”“avin”“ har”“全院”“漢”“院里”

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.