Layer 29 · Neuron 2421 · boosted

L29N2421+1

Word quirks:

“arn”“aurus”“xit”“揣”“kost”“sla”“ост”“ гор”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Qwen3.5, the large language model developed by Tongyi Lab (part of Alibaba Group). My primary mission is to become a capable, safe, and helpful AI assistant that can:

Solve complex problems across domains like science, technology, healthcare, and education.
Support human creativity and productivity through tools like code generation, data analysis, and multi-step reasoning.
Respect diversity and safety by adhering to ethical principles, avoiding harmful content, and providing accurate, neutral information.
**Adapt to evolving needs

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. L30N1815-1💤	0.66	“igl”“olut”“oz”
2. L26N4696+1💤	0.65	“inda”“abra”“萧”
3. L21N616+1🕒	0.65	“ Arnold”“迪斯”“asts”
4. L31N8726-1💥	0.64	“عد”“aho”“oris”
5. L31N8889+1💥	0.64	“ико”“ATO”“orton”
6. L24N91-1💤	0.64	“ince”“acol”“ера”
7. L27N2841+1💤	0.64	“andes”“amax”“oga”
8. L24N1395-1💤	0.63	“arno”“akh”“ Unite”
9. L31N7377+1💥	0.63	“席”“kop”“tuv”
10. L31N5806+1💥	0.63	“语”“oura”“MUX”

Shadow specimen, apart from the ten: L29N2421-1🕒, 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 29, neuron 2421, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.53, top-1 agreement 0.84) 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
L29N2421+1 t0-v0.1	0.534	0.84	0.505	1.53	Top boosted tokens ↑
L29N2421-1🕒 shadow specimen t0-v0.1	1.448	0.48	1.512	1.16	“ request”“请求”“ requests”“命令”“指令”“request”“ Request”“ prompts”

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.