Layer 29 · Neuron 5592 · boosted

L29N5592+1

Word quirks:

“瑶”“egi”“ayed”“曦”“azin”“昭”“猎”“syl”

💤 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 Qwen3.5, the latest large language model developed by Alibaba's Tongyi Lab. My mission is to empower human creativity and problem-solving by providing intelligent, accurate, and safe assistance across a wide range of tasks. Here's how I aim to deliver on this mission:

Core Objectives:

Ubiquitous Understanding & Reasoning

Master complex reasoning in mathematics, natural sciences, and logic.
Analyze multi-step problems with human-like intuition (e.g., debugging code, interpreting scientific papers).

**Precise Long-C

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. L30N4012+1🕒	0.78	“彦”“昌”“玮”
2. L26N4710-1💤	0.77	“rift”“صرية”“雪峰”
3. L31N632+1💥	0.76	“彪”“新社”“锦标”
4. L30N4592+1💤	0.76	“رخ”“飞来”“御”
5. L30N931+1💤	0.76	“僚”“fdc”“yal”
6. L30N759+1💤	0.76	“媛”“羡”“中航”
7. L30N6420+1🕒	0.76	“鹰”“載”“elen”
8. L26N2608-1💤	0.76	“凌”“太子”“信”
9. L29N7128-1💤	0.76	“rawn”“南山”“酰”
10. L29N3034+1💤	0.75	“朱”“承”“ngen”

Shadow specimen, apart from the ten: L29N5592-1💥, -0.05.

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 29, neuron 5592, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.27, top-1 agreement 0.79) 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
L29N5592+1 t0-v0.1	0.266	0.79	0.367	1.38	Top boosted tokens ↑
L29N5592-1💥 shadow specimen t0-v0.1	19.555	0.00	19.600	0.22	“ capacity”“ Capacity”“Capacity”“capacity”“ capacities”“容量”“_capacity”“容量的”

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.