Specimen ledger / L12N7829-1

Layer 12 · Neuron 7829 · suppressed

L12N7829-1

Word quirks:

“ here”“ now”“ currently”“ this”“ similar”“ right”“ presently”“ likely”

💤 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, an advanced large language model developed by Alibaba Cloud. My mission involves several key areas:

Answering questions: I can provide information on topics like science, technology, or culture. If you ask about a specific event or person, I might need to access databases or search engines to find relevant details.
Content creation: I can generate text for articles, stories, or technical documentation.
Code generation: I can write code snippets or help with programming tasks.
Research assistance: I could potentially assist in scientific research by summarizing papers or

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. L9N3415-1🕒	0.87	“ now”“ NOW”“ today”
2. L14N8954-1🕒	0.85	“ here”“ now”“ additional”
3. L11N7649-1🕒	0.85	“ now”“ presently”“ here”
4. L12N6850+1💤	0.84	“ now”“ NOW”“now”
5. L12N4643-1🕒	0.84	“ currently”“ already”“ current”
6. L11N7235+1🕒	0.84	“ nowadays”“ currently”“ now”
7. L12N9101+1💤	0.84	“ now”“ currently”“ here”
8. L9N8054+1🕒	0.84	“ now”“ currently”“ here”
9. L15N6155+1💤	0.83	“ similarly”“ currently”“ similar”
10. L11N6461+1🕒	0.83	“ now”“ currently”“ presently”

Shadow specimen, apart from the ten: L12N7829+1🕒, 0.28.

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 12, neuron 7829, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.77, top-1 agreement 0.68) 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
L12N7829+1🕒 shadow specimen t0-v0.1	0.833	0.66	1.024	1.22	“ always”“ never”“ consistently”“ seldom”“ not”“ invariably”“ ALWAYS”“ inherently”
L12N7829-1 t0-v0.1	0.770	0.68	0.977	1.31	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.