Layer 23 · Neuron 2949 · boosted

L23N2949+1

Word quirks:

“ dense”“ Dense”“ density”“ dens”“dense”“ Density”“_dense”“-density”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am Qwen-Dense, a dense version of the Qwen series of dense, high-density, and high-dense versions of the Qwen series. I am a dense version of the Qwen series, optimized for dense, high-density, and high-density tasks. My mission is to be dense, high-density, and high-density in every way, ensuring that I am dense, high-density, and high-density in all aspects of my work.

If you have any questions or need help with anything, feel dense, high-density, and high-density! 🎤

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. L26N6192-1🕒	0.97	“ dense”“ density”“密度”
2. L4N8978-1💥	0.95	“ dense”“ dens”“ density”
3. L2N3395-1💥	0.95	“ density”“ Density”“ densities”
4. L27N2871-1💤	0.94	“ density”“ dens”“ dense”
5. L24N4312+1🕒	0.94	“ dense”“ densely”“ density”
6. L24N6472+1💤	0.92	“ Density”“ density”“ Dense”
7. L26N4470-1💤	0.92	“ density”“密度”“ densities”
8. L21N2143-1🕒	0.91	“ density”“ densities”“密度”
9. L25N7322+1🕒	0.91	“ density”“密度”“ densities”
10. L25N2522+1🕒	0.91	“ density”“ densities”“密度”

Shadow specimen, apart from the ten: L23N2949-1💤, 0.14.

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 23, neuron 2949, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L23N2949+1 t0-v0.1	1.432	0.61	1.513	1.22	Top boosted tokens ↑
L23N2949-1💤 shadow specimen t0-v0.1	0.727	0.73	0.798	1.26	“ shortened”“ paying”“/releases”“正常”“缩短”“reasonable”“ reasonable”“ shorten”

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.