Experiments

Every experiment carries a registered hypothesis with an explicit falsification criterion (seven-field scientific block), mandatory controls and nulls, raw results, and an analysis. Negative results are kept. Micro-experiments A–H establish the noise floor, the correlational→causal survival rate, and the capture cost frontier before any large model is mapped.

expA_probe_reliability

Probe reliability — noise floor of linear probes across seeds and datasets

completed 1 result run

expB_localization_vs_diffusion

Localization vs diffusion — how concentrated is the signal supporting a capability

scaffolded

expC_causal_verification

Causal verification pipeline — correlational-to-causal survival rate

scaffolded

expD_cross_input_stability

Cross-input stability — which map types are input-invariant

scaffolded

expE_weight_geometry

Weight-space geometry — zero-forward-pass maps (SVD, effective rank, CKA)

scaffolded

expF_quantization_map_drift

Quantization map drift — how FP16/Q8/Q4/Q2 deform internal structure

scaffolded

expG_cross_model_alignment

Cross-model alignment — shared coordinate systems across model sizes

scaffolded

expH_capture_cost_frontier

Capture cost frontier — what can you map on which Mac (macOS-specific)

completed 3 result runs

candidate_01

Assigned by Phase 4 (`research/candidate_ranking.md`). Core gaps: G06 (+G08 anchor, G03/G09 extensions).

scaffolded

candidate_02

Assigned by Phase 4 (`research/candidate_ranking.md`). Core gaps: G01 (+G02, then G04/G05).

scaffolded

candidate_03

Assigned by Phase 4 (`research/candidate_ranking.md`). Core gaps: G17 (+G11/G12 as measurement layers).

scaffolded

candidate_04

Assigned by Phase 4 (`research/candidate_ranking.md`). Core gaps: G24 (+G09).

scaffolded