Layer 7 · Neuron 2507 · suppressed

L7N2507-1

Word quirks:

“ scholarly”“ scholars”“ scholar”“ scholarship”“ textual”“ phil”“ histori”“ texts”

🕒 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 Qian Zhongshu (1898–1992), a renowned Chinese sinologist and philologist. His most famous work in this field is the multi-volume series "Studies on the Classics" (Wang Guo) and his later comprehensive research project "The Study of the Classics" (Daxue).

His mission was to:

Reconstruct the original texts of ancient Chinese classics using rigorous philological methods.
Critically analyze the historical, linguistic, and cultural contexts of these texts.
**Challenge

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. L3N4594-1💥	0.96	“ textual”“ texts”“ scholarly”
2. L4N8602+1💥	0.89	“ phil”“ scholarly”“ scholars”
3. L4N8674+1💥	0.87	“ scholar”“ scholarly”“ scholars”
4. L5N1261+1💥	0.86	“ histori”“ scholars”“ historians”
5. L2N7297+1💥	0.84	“ scholar”“ Scholar”“ scholars”
6. L26N7168+1🕒	0.84	“sch”“ scholar”“ Scholar”
7. L5N5038-1🕒	0.83	“ scholarly”“ scholar”“ scholars”
8. L6N5124+1🕒	0.82	“ philosophical”“ philosophy”“ scholarly”
9. L24N849-1🕒	0.81	“sch”“ scholar”“ scholars”
10. L1N3636+1💥	0.78	“ scholar”“……”“ scholarly”

Shadow specimen, apart from the ten: L7N2507+1💤, -0.04.

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 7, neuron 2507, 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 0.49) 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
L7N2507+1💤 shadow specimen t0-v0.1	0.680	0.67	0.791	1.30	“ code”“ \"”“ impacting”“ system”“ triggering”“ '”“ impact”“ control”
L7N2507-1 t0-v0.1	2.052	0.49	2.435	1.86	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.