Layer 17 · Neuron 5646 · suppressed

L17N5646-1

Word quirks:

“ “”“ previously”“ earlier”“ ‘”“ prior”“...””“ another”“ first”

🕒 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 Qwen3.6, a large language model developed by the Qwen family of models. My primary mission is to understand and assist users in comprehending complex content and supporting their needs. Specifically, I focus on:

Enhancing my capabilities: I have been trained to better understand the context of a conversation, which helps me generate more coherent responses.
Improving my reasoning abilities: I can tackle more challenging tasks, such as multi-step reasoning and logical consistency.
Expanding my knowledge base: I support advanced technical applications, including code generation and understanding.

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. L11N2463+1💤	0.85	“ previously”“ previous”“ earlier”
2. L16N492-1🕒	0.84	“ earlier”“ **“”“ previously”
3. Phantom Memoir✅ L5N8933+1	0.83	“ earlier”“ ‘”“ later”
4. L14N1026+1🕒	0.82	“ previous”“ earlier”“ previously”
5. L14N2403+1💤	0.82	“ previously”“ previous”“<\|im_end\|>”
6. L7N9187-1💤	0.82	“⏎”“ previous”“ ‘”
7. L12N64-1🕒	0.81	“ earlier”“ already”“ “”
8. L13N3009+1💤	0.81	“<\|im_end\|>”“⏎”“ ‘”
9. L14N7199-1🕒	0.80	“ ‘”“ ⏎”“ previously”
10. L15N794+1🕒	0.79	“ earlier”“ previous”“ your”

Shadow specimen, apart from the ten: L17N5646+1💤, -0.03.

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 17, neuron 5646, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.64, top-1 agreement 0.71) 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
L17N5646+1💤 shadow specimen t0-v0.1	0.522	0.70	0.600	1.17	“ ideal”“ probably”“理想”“ maybe”“的理想”“ desired”“Ideal”“ Ideal”
L17N5646-1 t0-v0.1	0.638	0.71	0.836	1.31	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.