Layer 6 · Neuron 5689 · boosted

L6N5689+1

Word quirks:

“ cleanup”“ cleaned”“ cleaning”“ cleanliness”“ clean”“ tidy”“ Cleanup”“ cleans”

🕒 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 here. Please let me know what specific task you need help with (e.g., cleaning up the workspace, organizing files, or closing the session) so I can assist you in completing it.

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. L8N6205-1💤	0.97	“ cleanup”“ cleaned”“ cleanse”
2. L24N9166+1🕒	0.97	“ cleanup”“cleanup”“ Cleanup”
3. L8N829+1🕒	0.97	“ cleaned”“ cleaning”“ clean”
4. L9N4242-1🕒	0.96	“ cleaned”“ cleaning”“ cleanliness”
5. L7N8675-1🕒	0.96	“ cleaned”“ cleaning”“ cleanup”
6. L7N8731-1🕒	0.96	“ cleaning”“ cleaned”“ cleaner”
7. L8N5136-1🕒	0.96	“ cleaning”“ cleaner”“ cleans”
8. L7N3596-1💤	0.96	“ cleaning”“ cleaners”“ cleans”
9. L7N7720-1🕒	0.96	“ cleaning”“ cleanup”“ cleaned”
10. L5N7287+1🕒	0.95	“ cleaning”“ cleanup”“ cleaned”

Shadow specimen, apart from the ten: L6N5689-1🕒, 0.09.

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 6, neuron 5689, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.05, top-1 agreement 0.46) 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
L6N5689+1 t0-v0.1	2.050	0.46	2.467	2.06	Top boosted tokens ↑
L6N5689-1🕒 shadow specimen t0-v0.1	1.353	0.53	1.614	1.81	“ selection”“ selections”“ selected”“ select”“ combinations”“ choice”“ selecting”“ combos”

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.