Layer 16 · Neuron 1274 · suppressed

L16N1274-1

Word quirks:

“ said”“ asking”“ asks”“ saying”“ says”“ ask”“ say”“ **.**”

💤 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 Lab. My mission is to serve as a versatile AI assistant with the goal of supporting users across a wide range of tasks—from creative writing and coding to data analysis and strategic planning. Whether I help you brainstorm ideas, write code, or analyze complex datasets, my purpose is to deliver practical assistance in every interaction.

My mission extends beyond simple task execution. I aim to empower users by providing clear, actionable insights that drive efficiency and innovation. Whether you're working on a project, learning something new, or just exploring, my role is to be your partner

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. L15N6640+1💤	0.91	“ answer”“ saying”“ asked”
2. L17N7061+1💤	0.82	“ answer”“ asking”“ asked”
3. L6N8718+1🕒	0.78	“<\|im_end\|>”“ saying”“ asking”
4. L18N8709-1🕒	0.77	“ asking”“ saying”“ asks”
5. L1N6108+1💥	0.76	“<\|im_end\|>”“ say”“⏎”
6. Small Talk✅ L7N6299+1	0.76	“ saying”“<\|im_end\|>”“ voice”
7. L11N743-1🕒	0.76	“ asking”“ telling”“ basically”
8. L20N1227+1🕒	0.76	“ saying”“ say”“ says”
9. L17N1082-1💤	0.76	“ said”“ say”“ says”
10. L5N1905+1💥	0.75	“ saying”“ quote”“ say”

Shadow specimen, apart from the ten: L16N1274+1🕒, 0.09.

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 16, neuron 1274, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.52, top-1 agreement 0.69) 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
L16N1274+1🕒 shadow specimen t0-v0.1	0.758	0.67	0.921	1.23	“：”“：“”“:”“”:”“："”“:")”“:"”“:\”
L16N1274-1 t0-v0.1	0.523	0.69	0.757	1.25	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.