Specimen ledger / L12N4312-1

Layer 12 · Neuron 4312 · suppressed

L12N4312-1

Word quirks:

“ typical”“ usually”“ often”“ typically”“ usual”“ here”“ sometimes”“ relevant”

💤 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. Here's a brief overview of my identity and capabilities:

Identity: I am an AI assistant developed by Alibaba Cloud.
Mission: My goal is to provide users with high-quality, helpful, and safe interactions. I aim to assist with tasks that require creativity, reasoning, and practical utility.

Key Capabilities

Advanced Reasoning: I excel in complex problem-solving, logical analysis, and generating coherent content.
Multilingual Support: I can communicate fluently across multiple languages, including but not limited to Chinese, English, and others.

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. The Generalizer✅ L9N7460-1	0.97	“ usually”“ typically”“ sometimes”
2. L9N5744-1🕒	0.96	“ typically”“ sometimes”“ typical”
3. L13N5813-1🕒	0.96	“ here”“ often”“ typically”
4. L13N7031-1💤	0.96	“ typically”“ often”“ usually”
5. L10N8099+1🕒	0.95	“ sometimes”“ usually”“ usual”
6. L10N8560-1🕒	0.95	“ usually”“ often”“ sometimes”
7. L10N4952+1🕒	0.95	“ usually”“ often”“ sometimes”
8. L12N2355+1🕒	0.95	“ sometimes”“ often”“ typical”
9. L13N3347-1🕒	0.95	“ typical”“ often”“ typically”
10. L11N7468+1🕒	0.95	“ typical”“ typically”“ usually”

Shadow specimen, apart from the ten: L12N4312+1💤, -0.04.

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 4312, 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 0.67) 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
L12N4312+1💤 shadow specimen t0-v0.1	0.732	0.61	0.920	1.24	“ ⏎”“ ⏎⏎”“ *”“ ⏎ ⏎”“ **”“ ⏎”“ -->”“<\|im_end\|>”
L12N4312-1 t0-v0.1	0.639	0.67	0.852	1.15	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.