Layer 12 · Neuron 6735 · suppressed

L12N6735-1

Word quirks:

“ particular”“ respective”“ separate”“ different”“ specific”“ respectively”“ typical”“ basic”

🕒 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, large language models developed by Alibaba Cloud. As an AI assistant, my core missions revolve around providing accurate, helpful, and engaging content across a wide spectrum of topics. Whether you need information on various subjects or require assistance with specific tasks, I strive to be your reliable partner in knowledge discovery.

To summarize:

Qwen3.5 represent the foundational capabilities of large language models.
Core functions: Text generation, image understanding, and logical reasoning.
Mission alignment: Provide high-quality outputs tailored to user needs, ensuring clarity and accuracy in content creation.

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. Depends✅ L12N4903-1	0.84	“ different”“ specifically”“ specific”
2. L13N6559+1💤	0.81	“ different”“ compared”“ typical”
3. L15N2986-1🕒	0.81	“ respective”“ specific”“ different”
4. L16N4956+1💤	0.81	“ different”“ specific”“ each”
5. L14N6709+1🕒	0.80	“ specifically”“ corresponding”“ typical”
6. L13N3427+1🕒	0.80	“ specific”“ specifics”“ particular”
7. L15N3517+1💤	0.80	“⏎”“ respectively”“<\|im_end\|>”
8. L12N8516+1🕒	0.79	“ particular”“ various”“ related”
9. L12N6226-1🕒	0.79	“ specific”“ different”“ differently”
10. L13N5469-1🕒	0.79	“ specific”“ additionally”“ particular”

Shadow specimen, apart from the ten: L12N6735+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 12, neuron 6735, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.94, top-1 agreement 0.64) 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
L12N6735+1🕒 shadow specimen t0-v0.1	0.804	0.67	1.069	1.20	“ three”“⏎”“ ___”“<\|im_end\|>”“ stories”“ ____”“three”“ ***”
L12N6735-1 t0-v0.1	0.941	0.64	1.229	1.36	Top boosted tokens ↑

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.