Layer 18 · Neuron 6845 · boosted

L18N6845+1

Word quirks:

“ compared”“ normally”“ difference”“ similar”“ Reduced”“ comparison”“ Compared”“ similarly”

💤 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, a large language model with enhanced capabilities compared to earlier versions. My mission is to provide high-quality assistance across various domains, including:

Accurate Information Retrieval: Delivering precise answers based on my training data (as of 2026).
Technical Expertise: Assisting with coding, debugging, and explaining complex technical concepts.
Multilingual Support: Communicating in over 100 languages to broaden accessibility.
Contextual Understanding: Maintaining coherence across long conversations or multi-turn interactions.
Practical Application: Helping

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. L15N3303-1💤	0.90	“ differences”“ comparison”“ similar”
2. L13N4246+1💤	0.89	“ similar”“ compared”“ comparable”
3. L16N8977+1🕒	0.89	“ compared”“ similar”“ compares”
4. L14N183-1🕒	0.89	“ similar”“ same”“ similarly”
5. L17N5562+1🕒	0.88	“ compared”“ similar”“ comparable”
6. L15N7693-1🕒	0.88	“ compared”“ similar”“ than”
7. L13N8026+1🕒	0.87	“ similar”“ compared”“ focus”
8. L10N4742+1🕒	0.87	“ comparison”“ comparable”“ comparing”
9. L16N1685-1💤	0.87	“ similar”“ compared”“ similarly”
10. L11N2842+1🕒	0.87	“ similar”“ same”“ comparable”

Shadow specimen, apart from the ten: L18N6845-1🕒, 0.13.

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 18, neuron 6845, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.44, top-1 agreement 0.74) 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
L18N6845+1 t0-v0.1	0.437	0.74	0.565	1.12	Top boosted tokens ↑
L18N6845-1🕒 shadow specimen t0-v0.1	0.906	0.68	1.017	1.31	“ limitations”“ dissatisfaction”“limitations”“ satisfaction”“不能满足”“ limitation”“ Satisfaction”“无法满足”

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.