Layer 19 · Neuron 8095 · boosted

L19N8095+1

Word quirks:

“ competing”“竞争”“ compete”“ compet”“ competition”“ competitive”“ competit”“Compet”

🕒 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 Q3.5, an iterative model in the series of large models developed by or comparable to T3-Long or similar architectures (depending on the specific versioning). My mission is to provide high-quality information, reasoning capabilities, and support in various domains such as science, technology, creativity, and problem-solving. I aim to ensure that I can match or exceed performance in benchmarks while maintaining or surpassing my capabilities in long-context understanding, multi-modal reasoning, and content creation. If you have any specific topics or tasks, feel matched or outmatched in these areas. Let's see who wins! 🥇�

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. L25N6093+1🕒	0.98	“ competit”“ competition”“Compet”
2. Perfect Competition✅ L25N8855+1	0.98	“竞争”“ competitive”“ competition”
3. L21N5260+1💤	0.97	“ competitors”“ competitor”“ competitiva”
4. L23N7996+1💤	0.97	“ competitor”“ Compet”“竞争的”
5. L22N8771+1🕒	0.97	“ Compet”“竞争”“Compet”
6. L20N6480-1🕒	0.97	“ competitive”“competitive”“ competition”
7. L22N5386-1🕒	0.97	“ competition”“竞争”“ Competition”
8. L25N4972-1🕒	0.97	“ competition”“ Competition”“competition”
9. L17N5164+1🕒	0.97	“ competitor”“ competitors”“ competition”
10. L20N5180-1💤	0.97	“ competition”“ Competition”“ compete”

Shadow specimen, apart from the ten: L19N8095-1💤, 0.08.

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 19, neuron 8095, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.99, top-1 agreement 0.66) 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
L19N8095+1 t0-v0.1	0.994	0.66	1.231	1.38	Top boosted tokens ↑
L19N8095-1💤 shadow specimen t0-v0.1	0.452	0.78	0.544	1.21	“间隔”“ movement”“小城”“estre”“ distance”“ between”“之间的距离”“ rest”

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.