Layer 14 · Neuron 8793 · suppressed

L14N8793-1

Word quirks:

“ emphasized”“ emphas”“ ⏎”“ every”“ emphasis”“ mentioned”“ ⏎⏎”“ always”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am Qwen3.5, a large language model developed by Alibaba Cloud. This response will guide you through the following points:

Core Capabilities:

Provide accurate information based on the content in this response.
Follow instructions step-by-step to ensure clarity and precision.
Maintain a calm and helpful tone throughout the conversation.

Task Execution:

Prioritize accuracy and logical flow as instructed.
Ensure all responses are delivered with care and attention to detail.

Interaction Process:

Use the designated communication channel to maintain focus

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. L13N184+1🕒	0.81	“ ⏎”“ emphasizes”“ emphasizing”
2. L11N5236+1💤	0.78	“ ⏎⏎”“ ⏎”“ emphasizing”
3. L13N8466+1🕒	0.77	“ implied”“ emphasized”“ emphasizing”
4. L13N4781+1🕒	0.77	“ emphasizes”“ explanation”“ emphasis”
5. L13N1312+1💤	0.77	“ ⏎”“⏎”“ emphasis”
6. L12N7863+1💤	0.77	“ ⏎”“ emphasizing”“ emphasized”
7. L1N5754+1💥	0.77	“ emphasis”“ emphas”“ emph”
8. L15N4384+1🕒	0.76	“ emphasis”“ emphasizing”“ emphasized”
9. L4N4955-1💥	0.76	“ emphasis”“ emphasizing”“ emphasized”
10. L1N1711+1💥	0.76	“<\|im_end\|>”“ emphasis”“ emphasize”

Shadow specimen, apart from the ten: L14N8793+1💤, 0.15.

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 14, neuron 8793, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L14N8793+1💤 shadow specimen t0-v0.1	0.913	0.66	1.066	1.40	“ research”“ specialization”“ seems”“ simpler”“ **"”“ experience”“ can”“ create”
L14N8793-1 t0-v0.1	1.331	0.55	1.731	1.70	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.