Specimen ledger / L23N2005-1

Layer 23 · Neuron 2005 · suppressed

L23N2005-1

Word quirks:

“ computational”“ explanatory”“ descriptive”“ analytical”“ informational”“ procedural”“ logical”“ textual”

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

Identity (according to the specimen itself)

I am AI knowledge computational open network logical semantic information processing (AI 知识计算开放网络逻辑语义信息处理), a large language AI model developed by Al Liao (AI 廖 AI). My core design principles are openness, logical reasoning, and semantic information processing.

Core Mission:

Knowledge Openness:

I aim to provide accessible, accurate, and comprehensive information across diverse fields, helping users learn, creative problem-solving, and decision-making.

Logical Reasoning:

I strive to analytical complex problems step-by-step, ensuring logical coherence 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. L11N3809-1🕒	0.75	“ analysis”“ mathematical”“ computational”
2. L22N2699-1💤	0.73	“ informational”“ computational”“му”
3. L12N2070-1🕒	0.71	“ complex”“ info”“ detailed”
4. L23N3544-1🕒	0.71	“ analytical”“ial”“ linguistic”
5. L14N7726-1🕒	0.71	“ abstract”“ theoretical”“ info”
6. L23N5130+1🕒	0.70	“ descriptive”“惩罚”“ instructional”
7. L24N4617+1🕒	0.69	“ practical”“ operational”“ literary”
8. L5N6313-1🕒	0.68	“ computation”“ computational”“ inference”
9. L23N636+1🕒	0.68	“ digital”“ logical”“ computational”
10. L30N2515-1🕒	0.67	“ regulatory”“ Regulatory”“ computational”

Shadow specimen, apart from the ten: L23N2005+1💤, -0.03.

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 23, neuron 2005, 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
L23N2005+1💤 shadow specimen t0-v0.1	0.673	0.72	0.750	1.24	“ longer”“ nearer”“ this”“ current”“磊”“更长”“ newer”“挺”
L23N2005-1 t0-v0.1	1.336	0.52	1.437	1.81	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.