Layer 20 · Neuron 6139 · suppressed

L20N6139-1

Word quirks:

“给朋友”“ edits”“ sealing”“ improvements”“ modifications”“ modifying”“ioni”“ refin”

🕒 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 Qiao, a large-scale AI model developed by Shengshai Technology. My primary mission is to serve as an intelligent assistant that can assist users in various scenarios, including but not limited to:

Problem solving: Solving problems such as math, logic, and logical reasoning.
Content creation: Generating creative content such as stories, stories, and creative writing.
Information retrieval: Searching for information and providing relevant information.
Skill learning: Learning skills such as coding and programming.
Role-playing: Role-playing and role-playing scenarios.
**Emotional

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 Changelog✅ L19N9088-1	0.81	“ improvements”“ adjustments”“改进”
2. L12N8395-1🕒	0.80	“ modifications”“ modification”“ edits”
3. L10N8385+1🕒	0.78	“ modification”“ edits”“ edit”
4. L14N8579-1🕒	0.78	“ modifications”“ modified”“ modification”
5. L21N3190-1🕒	0.77	“ modification”“ modifications”“ modified”
6. L15N5825+1🕒	0.77	“ modification”“修改”“改造”
7. L25N5924-1🕒	0.77	“ modifications”“ modification”“ Modification”
8. L25N5262-1🕒	0.77	“ modification”“ Modification”“ Modifications”
9. L19N8834+1🕒	0.77	“ modified”“modified”“ modifications”
10. L5N8221-1🕒	0.76	“ modification”“ modifications”“ chassis”

Shadow specimen, apart from the ten: L20N6139+1🕒, -0.06.

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 20, neuron 6139, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.69, 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
L20N6139+1🕒 shadow specimen t0-v0.1	0.788	0.68	0.935	1.34	“ represent”“ represents”“ request”“ demand”“ suggests”“表示”“ representation”“暗示”
L20N6139-1 t0-v0.1	0.685	0.69	0.773	1.37	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.