Non-record: Technique Taxonomy — Tier List, Interaction Effects, and BPB Verification Tools

[robbiebusinessacc]

Non-record research submission

Synthesized taxonomy from 20+ merged PRs and 880+ open PRs. Curated reference, not primary research.

What is in this document (and not elsewhere):

• Technique tier list — S/A/B/C ranked with measured BPB deltas and source PRs
• Interaction effects matrix — which combos are sub-additive (e.g. XSA-all + TTT loses 0.006 from overlap)
• BPB verification checklist — formula checks, common errors, n-gram causality + distribution validity checks
• N-gram validity crisis — PR RFC: How to Clean Up All the Parameter Golf Submissions #886 findings, issue Illegal submissions megathread #677 organizer rulings, hash collision proof, proposed fixes
• FineWeb n-gram repetition rates (measured)
• CROWN-Q quantization technique (PR Record: CROWN-Q + Full GPTQ + SWA/EMA Blend — val_bpb 1.1186 (3-seed mean) #693)
• Negative results index — pointers to PR Non-record: Depth Recurrence in Parameter-Constrained Transformers — What Works, What Doesn't, and Why #363, Research: Why Novel Architectures Fail at 16MB — Throughput-Quantization Co-optimization #831, Non-record: Negative results — hardware alignment & quantization on 8xH100 #670, Non-record: Negative results — quantization algorithms, TTT, architecture, and self-generated GPTQ calibration study #756
• Parameter budget calculator with verified configs

Full writeup: See README.md

Related PRs: #363, #831, #670, #756, #693, #886, #579, #549