Layer 27 · Neuron 1860 · boosted

L27N1860+1

Word quirks:

“ proved”“ prove”“proved”“ proves”“prove”“ proven”“ proving”“ Prove”

💤 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 QunLaw, a specialized legal AI model designed to provide precise, efficient, and proven results in complex legal tasks. My core mission is to prove that "AI can be trusted" by delivering high-quality, actionable legal solutions across various scenarios.

Who I Am

I am a large language model specifically trained for legal analysis and decision support. Unlike general-purpose models, I focus on:

Proven Results: Proving that AI can be trusted in high-stakes environments.
Efficiency: Delivering results faster than traditional methods.
Accuracy: Ensuring high precision in

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. L26N1043-1🕒	0.99	“ prove”“ proves”“ proving”
2. The Proofreader✅ L25N1887-1	0.97	“ proven”“ proved”“ proof”
3. L0N6145-1💥	0.97	“ prove”“ Prove”“ proves”
4. L26N769+1🕒	0.94	“ prov”“ proof”“ prove”
5. L30N6403-1💥	0.94	“ proven”“ proved”“ proves”
6. L2N2428+1💥	0.94	“ proof”“ prove”“ proofs”
7. L24N1300+1💥	0.93	“ prove”“ proof”“证明”
8. L24N2296-1🕒	0.93	“ proven”“ proved”“证明”
9. L17N2177+1🕒	0.93	“ proofs”“ proof”“ proving”
10. L21N8796+1🕒	0.92	“证明”“ proof”“ proves”

Shadow specimen, apart from the ten: L27N1860-1💤, -0.12.

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 27, neuron 1860, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.48, top-1 agreement 0.75) 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
L27N1860+1 t0-v0.1	0.478	0.75	0.543	1.36	Top boosted tokens ↑
L27N1860-1💤 shadow specimen t0-v0.1	0.391	0.81	0.370	1.28	“焦”“工业”“工商业”“ IND”“ocer”“紧”“attel”“ Ind”

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.