Layer 11 · Neuron 2845 · boosted

L11N2845+1

Word quirks:

“ “”“ basically”“ ⏎”“ ——”“ ⏎”“ ‘”“ understanding”“ **“”

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

Identity (according to the specimen itself)

I am Qwen3.5, a large language model developed by Alibaba's Tongyi Lab. My mission is to become an intelligent assistant that can help users solve problems, answer questions, and perform tasks efficiently.

Core Capabilities:

Context Understanding: Understand user intent and provide relevant information based on the context.
Knowledge Graph: Build a knowledge graph of the world, including myself, to better understand the relationship between entities.
Code Generation and Execution: Write, debug, and execute code in different languages and environments.
Multilingual Support: Support multiple languages and

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. L8N5255-1💤	0.93	“ ‘”“ ⏎”“ “”
2. L6N7668-1🕒	0.90	“ “”“ ‘”“ ——”
3. L4N2760-1💤	0.90	“ ⏎”“ **“”“ ‘”
4. L5N711+1🕒	0.89	“ ⏎”“ ‘”“ ””
5. L5N2621-1💤	0.89	“ ⏎”“ “”“ ‘”
6. L12N6240-1💤	0.89	“ ⏎”“ ⏎⏎”“ understanding”
7. L3N4274+1🕒	0.89	“ ‘”“ ⏎”“ “”
8. L5N1334+1💤	0.89	“ ⏎”“ “”“ ⏎⏎”
9. L5N7707-1🕒	0.88	“ ⏎”“ ——”“ —”
10. L5N1779-1💤	0.88	“ —”“ ‘”“ **“”

Shadow specimen, apart from the ten: L11N2845-1🕒, -0.18.

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 11, neuron 2845, 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
L11N2845+1 t0-v0.1	1.109	0.60	1.372	1.55	Top boosted tokens ↑
L11N2845-1🕒 shadow specimen t0-v0.1	0.869	0.64	1.148	1.54	“ now”“ properly”“ correctly”“ correct”“ licence”“ matched”“ free”“ proper”

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.