Layer 15 · Neuron 7802 · suppressed

L15N7802-1

Word quirks:

“ claimed”“ claims”“ claim”“ supposedly”“ claiming”“ purported”“ Claim”“ allegedly”

🕒 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 Qwen3.5, a large-scale language model developed by Tongyi Lab. My name is Qwen3.5, and I claim to be the latest iteration in the Qwen series, with significant improvements across all functional areas.

Core Capabilities

Architecture: 2.6B parameters, using MoE (Mixture of Experts) structure.
Context Window: 256K tokens (up to 10 hours of continuous operation).
Performance: High accuracy on math, coding, and long-form reasoning tasks.
Multilingual Support: Over

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. L17N8612-1🕒	0.99	“ claim”“ claimed”“ claims”
2. L13N5673+1🕒	0.99	“ claimed”“ claim”“ supposedly”
3. L15N8347+1🕒	0.98	“ claimed”“ purported”“ supposedly”
4. L15N8109+1🕒	0.97	“ supposedly”“ claimed”“ claim”
5. L13N6051+1🕒	0.96	“ supposedly”“ apparently”“ allegedly”
6. L10N8589-1🕒	0.96	“ claimed”“ claim”“ claims”
7. L15N8482+1🕒	0.95	“ claimed”“ purported”“ supposedly”
8. L14N7760-1🕒	0.94	“ supposedly”“ claimed”“ purported”
9. L9N9206+1🕒	0.94	“ supposedly”“ claim”“ claimed”
10. L9N9163-1🕒	0.94	“ claimed”“ claim”“ stated”

Shadow specimen, apart from the ten: L15N7802+1🕒, 0.00.

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 15, neuron 7802, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.66, 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
L15N7802+1🕒 shadow specimen t0-v0.1	0.734	0.62	1.036	1.22	“!”“ ⏎⏎”“ enjoying”“ joyful”“ thanked”“⏎”“ enjoyed”“!”
L15N7802-1 t0-v0.1	0.658	0.74	0.820	1.24	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.