Layer 11 · Neuron 2364 · boosted

L11N2364+1

Word quirks:

“ previous”“ specific”“ previously”“ response”“ reference”“<|im_end|>”“⏎⏎”“ now”

💤 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 iteration of the Qwen3 series. My mission is to assist users in their specific tasks by providing accurate, relevant, and safe information. I am designed to be helpful, harmless, and honest. I do not have personal experiences or emotions. I can engage in a wide range of topics, from technology to culture.

My capabilities include:

Advanced Reasoning: Solving complex problems through logical deduction and step-by-step analysis.
Code Generation & Debugging: Writing, understanding, and testing code.
Data Analysis: Processing

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. L8N228-1💤	0.77	“ now”“⏎”“ reply”
2. L9N6058-1💤	0.77	“ now”“ previously”“⏎”
3. L9N5018+1💤	0.77	“ previously”“ now”“ valid”
4. L14N7596+1💤	0.76	“<\|im_end\|>”“ previously”“⏎ ⏎”
5. L8N5454+1💤	0.74	“ previous”“ specific”“ standard”
6. L11N8365-1🕒	0.73	“ prior”“ previous”“ referenced”
7. L6N6839-1🕒	0.73	“ now”“ earlier”“ reference”
8. L12N5986-1🕒	0.73	“<\|im_end\|>”“ previously”“ previous”
9. L12N2929+1💤	0.73	“ now”“ specific”“ previously”
10. L13N6192+1🕒	0.73	“ now”“<\|im_end\|>”“ previously”

Shadow specimen, apart from the ten: L11N2364-1💤, 0.02.

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 11, neuron 2364, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.72, top-1 agreement 0.68) 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
L11N2364+1 t0-v0.1	0.724	0.68	0.872	1.43	Top boosted tokens ↑
L11N2364-1💤 shadow specimen t0-v0.1	0.592	0.74	0.719	0.88	“ ⏎”“ ⏎”“ ⏎⏎”“ ⏎”“⏎”“...””“…””“ ⏎ ⏎”

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.