12 — KV slots replace the articulation prefix (Sprint 2 / S2.2)

Pre-registered 2026-07-02 (human-approved sprint setup, before implementation). Agents running this experiment MUST NOT edit this spec; deviations are reported as deviations. Depends on research/11 (S2.1 minimal organism) being merged.

Problem

S2.1's content channel is a token-burning articulation prefix: consolidated facts occupy visible prompt tokens on every forward pass. S1.3 (research/09, log 2026-07-01_s1_3_hf_kv_slot_injection.md) established that KV entries spliced immediately pre-blank are rank-for-rank equivalent to a text prefix at zero visible-token cost (and that 2× K/V norm scaling is catastrophic — scaling is out of scope here). This experiment retires the prefix from the minimal organism.

Hypothesis

H-KVCONTENT: replacing the S2.1 consolidated articulation prefix with the same content encoded once at consolidation time via HFDriver.encode_kv and spliced pre-blank via generate_with_kv at answer time preserves the loop's held-out behavior change with ZERO fact tokens in the visible prompt.

Conditions (identical protocol to research/11: same stage, split, seeds, K=N)

• C0 vanilla: bare HFDriver, no injection.
• C1 prefix organism: the S2.1 configuration exactly (upper anchor).
• C2 KV organism: identical organism, but consolidate() compiles the same content into KV entries (encoded once, cached on the organism) and answer() uses generate_with_kv with pre-blank splice. No set_articulation_prefix call anywhere in the C2 path. K/V norms unscaled (1×).

Prompt-token audit (mandatory): for every C2 probe, the visible prompt token count must equal C0's for the same probe. Any mismatch invalidates C2.

Primary metrics & acceptance

1. kv_effect_delta = mean over seeds of [C2 holdout accuracy − C0 holdout accuracy] at K=N.
2. kv_parity_delta = mean over seeds of [C2 − C1 holdout accuracy] at K=N.

• Accept H-KVCONTENT: kv_effect_delta > 0 with 95% CI excluding 0, AND kv_parity_delta >= −0.05 (non-inferiority vs the prefix).
• Refute: either fails.

Validity gates: research/11's vanilla gate applies; if research/11 REFUTED H-LOOP (C1 itself produces no effect), this experiment is reported as BLOCKED, not run to a fake verdict.

Pre-registered secondary analyses (exploratory only)

1. Per-answer latency: C2 vs C1 vs C0 (ms, mean over holdout probes).
2. KV bytes vs prefix bytes in memory_bytes accounting.
3. Rank tables (target-token rank at blank) on the 3 lane_02 facts under the organism's consolidated state, C1 vs C2 — continuity check against S1.3.

Reporting

Per-seed table for C0/C1/C2; prompt-token audit result; model id; exact commands; log to experiments_logs/2026-07-02_s2_2_kv_content_path.md quoting this spec.

Amendment 2026-07-02 (before any primary verdict was drawn)

The pre-registered split call split_probes(stage, fraction=0.3, salt="v2") was discovered to be degenerate on stage 0: an unlucky hash assigned all 8 probes to dev, 0 to holdout — a state validate_split itself defines as an ERROR. The first S2.2 run executed against this empty holdout and is recorded as INVALID (instrument failure); its 0/0 "REFUTE" is void and carries no evidential weight for or against H-KVCONTENT.

Repair (instrument-level, not spec-level): split_probes now guarantees a non-empty holdout for every stage by promoting the lowest-force-hash probes to ceil(fraction × total) when thresholding yields none (stage 0: 3 holdout probes). No previously non-empty split is altered (locked by regression test test_split_guarantee_never_alters_nonempty_holdouts), so all previously logged numbers remain comparable. The spec's split call, salt, and fraction are unchanged. Amendment applies identically to research/11, 12, and 13.