Poor Things
Specimen ledger / L2N1903+1

Layer 2 · Neuron 1903 · boosted

L2N1903+1

Word quirks:

“<|im_end|>”“⏎ ⏎”“⏎”“⏎ ⏎”“?\”“...””“?””“.\"”

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

Identity (according to the specimen itself)

The reply didn't hold together

Line 2014-08-09 [00:00:00:00:00:00:00:00:00:00:00]
</think>

Line 2014-08-09 [00:00:00:00:00:00:00:00]
</think>

Line 2014-08-09 [00:00:00:00:00:00:00:00]
</think>

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

SpecimenSimilarityWord quirks
1. Hamster Wheel✅ L4N63+10.97“<|im_end|>”“⏎”“...””
2. L2N7683+1🕒0.97“<|im_end|>”“⏎”“⏎ ⏎”
3. L2N8613+1🕒0.97“<|im_end|>”“⏎”“...””
4. L1N1006-1💥0.96“<|im_end|>”“.\"”“⏎ ⏎”
5. L2N4172-1🕒0.96“<|im_end|>”“⏎”“⏎ ⏎”
6. L2N3510+1💥0.96“<|im_end|>”“⏎”“⏎ ⏎”
7. L3N529-1🕒0.95“<|im_end|>”“?\”“⏎”
8. L2N1822+1🕒0.95“⏎”“<|im_end|>”“...””
9. L2N1741-1🕒0.95“<|im_end|>”“?\”“⏎ ⏎”
10. L2N2049+1🕒0.95“<|im_end|>”“?\”“⏎”

Shadow specimen, apart from the ten: L2N1903-1💥, 0.81.

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 1903, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.79, top-1 agreement 0.33) 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.

SettingKL divergenceTop-1 agreementΔNLLEntropySteer lens
L2N1903+1
t0-v0.1		 2.7900.333.2152.47 Top boosted tokens ↑
L2N1903-1💥 shadow specimen
t0-v0.1		 3.6880.264.1712.91 “<|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.