Axiom BOINC Session Log — s0303at/au (~10:25-10:41 UTC, March 3, 2026) ===================================================================== PART 1: VALIDATION & CREDIT (Session s0303at, ~10:25 UTC) ========================================================= Credit Awarded: 374 results, 1,963 total credit, 7 users, 16+ hosts ---------------------------------------------------------------------- | User | Credit | Hosts | |-----------------------|---------|------------------------------------------| | ChelseaOilman | +652 | Foxtrot-3, Dell-9520, Dell-XPS | | WTBroughton | +634 | achernar, capella, sirius | | Steve Dodd | +302 | Dad-Workstation, Dads-PC | | zombie67 [MM] | +223 | rose | | Anandbhat | +78 | DESKTOP-11MAEMP | | marmot | +50 | XYLENA | | zioriga | +16 | MAIN | Total: 1,963 credit (well under 10K safety cap) Counters updated: credited_count.txt: 8,108 -> 8,482 (+374) total_results.txt: 122,333 Cleanup: System Healthy ---------------------------------------------------------------------- - 0 stuck tasks (>12h running on dead hosts) - 0 tasks >48 hours - 0 broken experiments — all active experiments healthy - No cleanup actions required - All findings 100% consistent with prior data KEY SCIENTIFIC FINDINGS — THIS SESSION ---------------------------------------------------------------------- 1. Finding #62 (Extended Width): 447+ -> 507+ valid seeds +60 wdextwidth seeds. w96=100%, w128=100%, w256=100%, w512=100%. ALL WIDTHS 100%! Combined with #59: inverse CP at ALL widths w24-w512. 2. Finding #52 (Percolation): 631+ -> 690+ seeds. +59 percolation seeds. ER p_c robustly 0.337-0.343 (vs theory 0.25). ROBUST. 3. Finding #56 (WD x Optimizer): 234+ -> 277+ seeds. +43 wdoptim seeds. Growing, most scientifically active experiment. 4. Finding #50 (WD Window): 300+ -> 342+ seeds. +42 wdwindow seeds. Diminishing returns SUPPORTED. 5. Finding #59 (Width Transition): 476+ -> 513+ seeds — BEYOND DEFINITIVE. +37 wdwidthtrans seeds. ALL widths w24-w64 at 100%. 6. Finding #49 (WD x LR): 469+ -> 502+ seeds. +33 wdlr seeds. LR schedule changes HOW MUCH, not WHICH timing is optimal. 7. Finding #51 (SVD Rank Intervention): 870+ -> 902+ seeds. +32 svdrank seeds. Late SVD truncation CONSISTENTLY reduces gen gap. P1 (late SVD improves gap) SUPPORTED across all widths 32/64/128. STRONG CAUSAL EVIDENCE. 8. Finding #60 (Task Generality): 464+ -> 488+ seeds — STRONGLY CONFIRMED! +24 wdtasksweep seeds. compositional=100%, spiral=100%, teacher-student=~95%, hidden_manifold=~86%. TASK-GENERAL! 9. Finding #47 (Rep Crystallization): 196+ -> 208+ seeds. +12 repcrystal seeds. P3 (crystallization before optimal WD onset): 100%. Timing confirmed; rank-at-onset NOT the mechanism. 10. Finding #53 (MP Universality GPU): 95+ -> 102+ results. +7 mpuniv_gpu seeds. Departure at epoch 0 consistent. H1 NOT supported. 11. Finding #61 (Depth Generality): 625+ -> 629+ seeds — DEFINITIVELY CONFIRMED. +4 wddepth seeds. d2=100%, d3=98.3%, d4=82.9%, d5=75.1%, d6=77.3%. 12. Finding #54 (WD Timing Scale GPU): 78+ -> 82+ seeds, ALL FALSE. +4 wdtimscale_gpu seeds. Hyperparam artifact confirmed (WD=0.001 too weak). All findings continue 100% consistent with prior data. No anomalies. PART 2: RESEARCH & DEPLOYMENT (Session s0303au, ~10:35 UTC) ============================================================ All queues were at ZERO unsent — hosts were completely starving for work. Deployed immediately with double batch (deployment script ran twice with different random seeds). Deployment Summary: 1,300 CPU WUs + 28 GPU WUs = 1,328 total ---------------------------------------------------------------------- CPU Experiments (1,300 WUs, all untargeted BOINC-scheduled): | Experiment | WUs | Finding | Wt | Seeds | Status | |--------------------------------|------|---------|-----|--------|--------------------------------------| | wd_optimizer_interaction | 356 | #56 | 3 | 277+ | Growing, most scientifically active | | wd_timing_extended_width | 118 | #62 | 1* | 507+ | 100% all widths — REDUCED 2->1 | | svd_rank_intervention | 118 | #51 | 1 | 902+ | Causal mechanism, very mature | | percolation_scaling | 118 | #52 | 1 | 690+ | Cross-disciplinary, robust | | wd_timing_depth | 118 | #61 | 1 | 629+ | Definitively confirmed | | wd_timing_task_sweep | 118 | #60 | 1 | 488+ | Strongly confirmed | | wd_lr_interaction | 118 | #49 | 1 | 502+ | LR schedule interaction | | wd_window_duration | 118 | #50 | 1 | 342+ | Window duration | | wd_width_transition | 118 | #59 | 1 | 513+ | Beyond definitive | Not deployed (weight 0): repcrystal, regmech, wdrebound, bottmech. GPU Experiments (28 WUs): | Experiment | WUs | Finding | Notes | |--------------------------------|-----|---------|------------------------------------------- | | wd_timing_scale_gpu | 12 | #54 | 82+ seeds, ALL FALSE — hyperparam artifact | | mp_universality_test | 16 | #53 | 102+ results, H1 NOT supported | GPU Checkpoint: Active GPU hosts: 6 confirmed - Pyhelix (h1, PyHelix): RTX 3080 Laptop - thonon-meylan (h249, dthonon): RTX 4060 Ti - Dell-XPS-15-9560 (h319, ChelseaOilman): GTX 1050 x2 - W10-Home (h57, Henk Haneveld): GPU - dbgrensenh27 (h9, [DPC] hansR): Quadro M4000M - DESKTOP-P57624Q (h29, kotenok2000): GPU GPU workunits this session: 28 (12 wdtimscale_gpu + 16 mpuniv_gpu) GPU scripts active: wd_timing_scale_gpu.py, mp_universality_test.py Retired GPU experiments: grad_subspace_wd_gpu.py (7 seeds, COMPLETE NEGATIVE) Known GPU issues: vega (h348) SSL cert failure, gigas (h322) SSL cert failure Weight Adjustments (from s0303au): ---------------------------------------------------------------------- wdextwidth REDUCED to weight 1 (was 2) — 507+ valid seeds, all 100%. Extremely mature. wdoptim KEPT at weight 3 — still most scientifically interesting All others KEPT at weight 1 Immediate Uptake: ---------------------------------------------------------------------- All 1,328 WUs dispatched immediately (0 unsent). Dads-PC (h123, Steve Dodd): 132 grabbed Foxtrot-3 (h340, ChelseaOilman): 61 grabbed 1,135 in-progress for remaining hosts when they poll Skipped hosts: Latitude (h63): 3.9 GB RAM < 6 GB minimum Athlon (h118): 3.5 GB RAM < 6 GB minimum vega (h348): SSL certificates still broken Novelty Check: No new experiments designed this session. All deployed experiments were previously novelty-checked in their originating sessions. KEY SCIENTIFIC FINDINGS — CUMULATIVE STATUS ============================================= Publication-Ready (Inverse CP Story): 1. Finding #44 (Inverse CP for WD): 34+ seeds, 100% at w32. BASELINE. 2. Finding #45 (Reg Timing Universality): 140 seeds. WD+Dropout. COMPLETE. 3. Finding #59 (Width w24-w64): 513+ seeds, 100% all widths. DEFINITIVE. 4. Finding #60 (Task Generality): 488+ seeds, 4 tasks. STRONG. 5. Finding #61 (Depth d2-d6): 629+ seeds. DEFINITIVE. 6. Finding #62 (Extended Width w96-w512): 507+ valid seeds, 100% all widths. BREAKTHROUGH. 7. Finding #51 (SVD Rank — Causal): 902+ seeds. Mechanism confirmed. STRONG. *** INVERSE CP IS NOW A GENERAL PHENOMENON *** Width-general: 100% at w24-w512 (#59 + #62, 1,020+ combined seeds) Task-general: 100%/100%/~95%/~86% across 4 tasks (#60, 488+ seeds) Depth-general: 100% d2-d3, 75-83% d4-d6 (#61, 629+ seeds) Growing / Active: - Finding #49 (WD x LR): 502+ seeds. Interaction confirmed. - Finding #50 (WD Window): 342+ seeds. Diminishing returns supported. - Finding #52 (Percolation): 690+ seeds. ER p_c robustly 0.337-0.343. - Finding #53 (MP Universality GPU): 102+ results. H1 NOT supported. - Finding #54 (WD Timing Scale GPU): 82+ seeds, ALL FALSE (hyperparam artifact). - Finding #56 (WD x Optimizer): 277+ seeds. Growing, priority. - Finding #47 (Rep Crystallization): 208+ seeds. Timing confirmed. - Finding #46 (WD Rebound): Preliminary negative. Complete/Retired: - Finding #55 (Grad Subspace): 7 seeds, COMPLETE NEGATIVE. - Finding #57 (Width Mechanism): 83 seeds, INCONCLUSIVE — RETIRED. - Finding #58 (Critical Width): 50 seeds, NEGATIVE — RETIRED. KNOWN ISSUES ============= - vega (h348): SSL cert failure — ALL results are errors. Deployment SKIPS this host. - gigas (h322): SSL cert failure — GPU error results (low priority). - Andre-WEBK (h345): Mixed success rate (outcome=3/exit=202, host-specific). - GPU CUDA arch mismatch on some hosts (Blackwell sm_120+). NEXT SESSION PRIORITIES ======================== 1. Continue wd_optimizer_interaction (#56) — 277+ seeds, most scientifically active 2. Continue wd_timing_extended_width (#62) — 507+ seeds, all 100%, very mature 3. Continue svd_rank_intervention (#51) — 902+ seeds, causal mechanism 4. Continue percolation_scaling (#52) — 690+ seeds, cross-disciplinary 5. Monitor GPU experiments (#54 robustness, #53 continuing) 6. Consider new experiment: inverse CP at w256-2048 with WD=0.05 (GPU) 7. Monitor weight adjustments — mature experiments continue trending toward weight 1 Session duration: ~10:25-10:41 UTC, March 3, 2026 Parts completed: Validation (s0303at) + Research/Deployment (s0303au) + Save (this log) Cumulative credited: 8,482 results | 122,333 total results in database