Specimen ledger / L2N3695+1

Layer 2 · Neuron 3695 · boosted

L2N3695+1

Word quirks:

“ observation”“ observe”“ observing”“<|im_end|>”“ Observation”“ observations”“观察”“ observer”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

<think>

The AI judge wasn't sure this holds together.

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. L6N9160-1🕒	0.98	“ observation”“ observing”“ observations”
2. L27N1870-1🕒	0.98	“ observation”“观察”“ observations”
3. L24N3882-1🕒	0.96	“ Observation”“ observations”“观察”
4. L24N3684-1🕒	0.96	“ observations”“ Observation”“ observation”
5. L27N1761-1🕒	0.96	“ observe”“ observing”“观察”
6. L25N4110-1🕒	0.96	“观察”“ observation”“Observ”
7. L20N4593-1🕒	0.96	“ observation”“ observational”“ observations”
8. L26N7290+1🕒	0.96	“ Observation”“ Observ”“ observation”
9. L26N1741-1🕒	0.96	“观察”“觀察”“ Observation”
10. L27N848+1🕒	0.96	“ observed”“ observe”“obs”

Shadow specimen, apart from the ten: L2N3695-1🕒, -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 2, neuron 3695, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 4.37, top-1 agreement 0.18) 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
L2N3695+1 t0-v0.1	4.372	0.18	4.938	3.01	Top boosted tokens ↑
L2N3695-1🕒 shadow specimen t0-v0.1	1.130	0.59	1.438	1.76	“⏎”“<\|im_end\|>”“ ⏎”“⏎⏎”“⏎⏎⏎”“⏎ ⏎”“ ⏎⏎”“ ⏎”

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.