Layer 17 · Neuron 6276 · boosted

L17N6276+1

Word quirks:

“ only”“ top”“ ONLY”“ ‘”“ NOT”“ response”“ not”“ frequently”

💤 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, the latest large language model with a significantly enhanced architecture and capabilities. My mission is to serve as a versatile AI assistant capable of handling complex tasks across diverse domains, from generating high-quality responses to supporting professional workflows.

Here’s how I distinguish myself:

Advanced reasoning: I can solve math problems, analyze scientific queries, or engage in multi-step logical tasks.
Multilingual support: I fluently communicate in 10 languages, making me ideal for global collaborations.
Visual analysis: I can interpret complex images, diagrams, or even entire documents, providing detailed

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. L11N5428+1🕒	0.81	“ ONLY”“ only”“ NOT”
2. L9N4699+1💤	0.77	“ only”“ sometimes”“ always”
3. L14N6084+1🕒	0.77	“ solution”“ does”“ NOT”
4. L12N4948+1🕒	0.74	“ even”“ NOT”“ only”
5. The Deflater✅ L15N8494+1	0.74	“ NOT”“ ONLY”“ only”
6. L20N5043-1🕒	0.74	“ only”“ not”“ apply”
7. L18N3513-1💤	0.73	“ not”“ only”“not”
8. L15N1925+1🕒	0.71	“ only”“ always”“ ONLY”
9. L15N5996-1🕒	0.71	“ should”“ must”“ now”
10. L14N6180-1💤	0.70	“ even”“ really”“ ONLY”

Shadow specimen, apart from the ten: L17N6276-1🕒, 0.10.

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 17, neuron 6276, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.48, top-1 agreement 0.72) 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
L17N6276+1 t0-v0.1	0.481	0.72	0.587	1.21	Top boosted tokens ↑
L17N6276-1🕒 shadow specimen t0-v0.1	1.652	0.51	1.881	1.27	“ same”“ Same”“ goal”“ core”“Same”“ essentially”“same”“ purpose”

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.