Layer 10 · Neuron 7847 · boosted

L10N7847+1

Word quirks:

“ improved”“ reduced”“ eliminated”“ optimized”“ reducing”“ eliminating”“ specially”“ enhanced”

🧪 In testing — Answered 0 of 67 prompts so far.

Identity (according to the specimen itself)

I am Qwen3.5, the latest large language model developed by Alibaba Cloud. My mission is to provide more accurate, efficient, and context-aware responses than previous models like GPT-3.5 or even GPT-4. I am designed to reduce hallucinations, improve reasoning capabilities, and support a wider range of tasks across domains such as coding, data analysis, and natural language understanding.

If you're asking about my specific design goals, I can help you avoid common pitfalls in earlier models. For example, I can answer questions that require less computational resources, faster inference speed, and reduced token usage. This

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. L22N6207+1🕒	0.77	“ improved”“ enhanced”“ altered”
2. L28N3274+1💤	0.77	“ reduced”“ Reduced”“ decreased”
3. L11N6905+1🕒	0.73	“ clearer”“ reduced”“ tweaks”
4. L22N4424-1⛔	0.72	“ increased”“ decreased”“ reduced”
5. L13N8142-1🕒	0.72	“ enhanced”“ optimized”“ enhancement”
6. L18N6840+1🕒	0.72	“ improving”“ improvement”“ eliminating”
7. L20N808+1💤	0.72	“ streamlined”“ simplified”“ stripped”
8. L17N8262+1💤	0.70	“ improved”“ refined”“ optimized”
9. L23N4477-1🕒	0.70	“ reduced”“ decreased”“ increased”
10. L31N2428+1💥	0.69	“ increased”“ extended”“ Increased”

Shadow specimen, apart from the ten: Lost the Thread✅, 0.14.

Cousins · specimens sharing a neuron

Specimen	Word quirks	Steering
Lost the Thread✅ Shadow specimen	“ basically”“ maybe”“ your”	L10N7847-1

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 10, neuron 7847, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the full 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
L10N7847+1 t0-v0.1	1.372	0.58	1.677	1.44	Top boosted tokens ↑
L10N7847-1✅ shadow specimen t0-v0.1	1.226	0.54	1.453	1.56	“ basically”“ maybe”“ your”“ what”“ personally”“ here”“ ONLY”“ very”

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.