Layer 24 · Neuron 722 · suppressed

L24N722-1

Word quirks:

“ transmission”“transaction”“传输”“场外”“交易”“ transaction”“балан”“transactions”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I'm Cristo, a specialized AI model designed to assist users with complex reasoning, scientific analysis, and creative tasks. My core mission is to provide high-quality, accurate, and actionable solutions across diverse domains—whether it's solving mathematical equations, analyzing scientific data, generating code, or crafting creative stories.

Key Focus Areas:

Deep Reasoning: Tackle multi-step problems in physics, chemistry, or advanced mathematics with rigorous logic.
Scientious Exploration: Analyze research papers, simulate experiments, or explain complex theories (e.g., quantum mechanics or climate modeling).
**Code

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. L26N4628-1🕒	0.89	“交易”“ transaction”“ Transaction”
2. L25N6537-1🕒	0.89	“ transaction”“transaction”“ Transaction”
3. L22N6886+1🕒	0.88	“交易”“的交易”“ transaction”
4. L25N4230+1🕒	0.87	“ transaction”“交易”“transaction”
5. L25N4215+1🕒	0.87	“ transaction”“交易”“transaction”
6. L26N2451-1🕒	0.86	“ transaction”“transaction”“ transactions”
7. L25N4796+1🕒	0.86	“ transaction”“transaction”“ transactions”
8. L30N8046+1💥	0.86	“ transaction”“ Transaction”“ transactions”
9. L29N4189-1💤	0.85	“ transaction”“transaction”“Transaction”
10. L1N6199+1💥	0.84	“交易”“ Transaction”“ transaction”

Shadow specimen, apart from the ten: L24N722+1🕒, -0.10.

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 24, neuron 722, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.79, 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
L24N722+1🕒 shadow specimen t0-v0.1	1.218	0.60	1.381	1.18	“ speak”“ speaks”“ speaking”“ Speak”“ spoke”“Speak”“spe”“Speaking”
L24N722-1 t0-v0.1	0.789	0.71	0.768	1.26	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.