Specimen ledger / L10N2021-1

Layer 10 · Neuron 2021 · suppressed

L10N2021-1

Word quirks:

“⏎”“!”“!:”“ now”“ ______”“!\”“ !”“ ____”

💤 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, a large-scale language model designed to assist with complex tasks, provide in-depth analysis, and generate creative content. My capabilities include understanding natural language, performing logical reasoning, and handling multi-step problems across various domains. I can also translate text between different languages and summarize long documents into concise insights.

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. L6N7611-1🕒	0.91	“!"”“ ____”“!”
2. L2N1312-1🕒	0.89	“<\|im_end\|>”“⏎”“ !”
3. L5N1471-1🕒	0.89	“⏎”“<\|im_end\|>”“ !”
4. L6N7146+1💤	0.89	“⏎”“!”“ _____”
5. L5N1955-1🕒	0.89	“<\|im_end\|>”“⏎”“!”
6. L11N1654-1🕒	0.88	“ ⏎⏎”“ ⏎”“!”
7. L6N464+1🕒	0.88	“ _____”“ !”“!”
8. L5N5197+1🕒	0.87	“!”“ !”“"!”
9. L8N477+1🕒	0.87	“ now”“ !”“!”
10. L2N7181+1🕒	0.87	“<\|im_end\|>”“⏎”“!\”

Shadow specimen, apart from the ten: L10N2021+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 10, neuron 2021, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.63, top-1 agreement 0.70) 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
L10N2021+1💤 shadow specimen t0-v0.1	0.628	0.69	0.730	1.07	“ referenced”“ actually”“ --”“ this”“ -”“ reference”“ especially”“ perfect”
L10N2021-1 t0-v0.1	0.626	0.70	0.830	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.