Specimen ledger / L11N1954-1

Layer 11 · Neuron 1954 · suppressed

L11N1954-1

Word quirks:

“ actually”“ specific”“ specifically”“ interesting”“ actual”“ sometimes”“ specifics”“ mentioned”

💤 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 developed by Alibaba Cloud. My capabilities extend far beyond basic interactions, covering a wide range of topics and tasks. Here's a breakdown of my key features:

Qwen3.5: I am the latest version in the Qwen series, featuring significant upgrades over previous versions like Qwen or Qwen1.
Context Window: I support extremely long documents, with context windows reaching up to 32,768 tokens for some use cases.
Code Capabilities: I can generate code snippets for various purposes, including web development and

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. L14N8658+1🕒	0.88	“ specifics”“ now”“ specific”
2. L13N2792-1🕒	0.88	“ specific”“ specifically”“ particular”
3. L14N7529+1💤	0.87	“ specific”“ specifically”“ actual”
4. L10N3862+1💤	0.87	“ specifically”“ specific”“ actually”
5. L13N4301-1🕒	0.87	“ sometimes”“ specific”“ specifics”
6. L10N1609+1💤	0.87	“ actually”“ “”“ specifics”
7. L15N7582+1🕒	0.86	“ actually”“ specific”“ generally”
8. L19N218+1🕒	0.85	“ actually”“ specifically”“ actual”
9. L15N44+1💤	0.85	“ specific”“ actually”“ specifics”
10. L15N1410-1🕒	0.85	“ especially”“ specifics”“ specifically”

Shadow specimen, apart from the ten: L11N1954+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 11, neuron 1954, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.78, top-1 agreement 0.62) 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
L11N1954+1🕒 shadow specimen t0-v0.1	1.064	0.62	1.309	1.39	“ showing”“ visual”“ shown”“ show”“ depiction”“ shows”“ depicted”“ depicting”
L11N1954-1 t0-v0.1	0.782	0.62	0.974	1.33	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.