Layer 20 · Neuron 2646 · boosted

L20N2646+1

Word quirks:

“ lift”“ Lift”“ lifting”“ lifts”“ lifted”“lift”“lifting”“选择”

🕒 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 Qwen2, a large language model developed by Completions. My mission is to help you make decisions and complete tasks by providing intelligent guidance and solutions. I can assist with:

Problem-solving: Choosing the right path for your specific situation.
Task completion: Helping you complete tasks efficiently.
Decision-making: Providing insights to guide your choices.
Task extension: Extending your capabilities beyond what you've already done.

How can I help you today? 😊

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. L30N2409+1💥	0.98	“ lift”“ lifted”“lift”
2. L24N4290+1🕒	0.98	“ lifted”“ lifting”“ Lift”
3. L25N7435-1💥	0.98	“ lift”“ lifting”“ Lift”
4. The Stuck Elevator✅ L26N702+1	0.97	“ lift”“ lifting”“ Lift”
5. L2N7425+1💥	0.96	“<\|im_end\|>”“ lift”“ lifting”
6. L25N6954+1🕒	0.88	“ lifting”“ lift”“ lifted”
7. L22N6186-1🕒	0.87	“ lifting”“ lifts”“ lift”
8. L25N8977-1💤	0.87	“ lift”“ Lift”“lift”
9. L29N7241-1🕒	0.84	“ lift”“lift”“ chair”
10. L20N4749+1🕒	0.83	“提升”“ elev”“ lift”

Shadow specimen, apart from the ten: L20N2646-1🕒, 0.17.

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 20, neuron 2646, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.85, top-1 agreement 0.71) 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
L20N2646+1 t0-v0.1	0.853	0.71	0.921	1.31	Top boosted tokens ↑
L20N2646-1🕒 shadow specimen t0-v0.1	0.748	0.68	0.905	1.14	“ safety”“ Safety”“主导”“Safety”“康健”“egel”“-dominated”“ reality”

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.