world-model-qa-mario

Archive of trained weights, configs and small evaluation artifacts for the Mario experiments (DIAMOND w-series, o-series, MAF m3 level scaling) of the private world-model-qa repository (Miraiyochi, inc.).

  • Source code / experiment READMEs: world-model-qa commit 5fbb2cbdc97f16c3229c45e951802745fe6fdaee (each experiment lives in experiments/<name>/; its README is the canonical record).
  • Moved here from Runpod network volumes on 2026-11-09 UTC; every file was SHA-256 hashed on the volume before upload and re-checked against the Hub LFS hash (SHA256SUMS.volume).
  • Excluded on purpose: optimizer-bearing intermediate checkpoints, public datasets, regenerable latent caches, Python environments.
  • Public archive (made public 2026-11-09). The UniReps 2026 H3 weights live in the public repo Miraiyochi/hierarchical-latent-subgoals; identical files are not duplicated here.

Notes

  • The Mario M1 corpus, M1b human play and the M2 DINOv3 latents from volume fhkh85lzmf are in the public dataset Miraiyochi/world-model-qa-mario-data (frames zstd-compressed, SHA-256 checked after decompression), not here.
  • The M2 flat chunk-scaling weights on volume fhkh85lzmf were not archived (PI decision, 2026-11-09).

Experiment → experiments/<name>/ → files in this repo

experiment (experiments/…) folder in this repo files size source volume
(wm-diamond legacy run logs) wm-diamond_records/ 17 0.00 GB uszvmpyjnb
hilewm_o10_issue_gate_v2 wm-diamond_results/hilewm_o10_issue_gate_v2/ 13 0.00 GB uszvmpyjnb
hilewm_o10_terminal_escalation wm-diamond_results/hilewm_o10_terminal_escalation/ 12 0.00 GB uszvmpyjnb
mario_maf_m1_corpus wm-diamond_results/mario_maf_m1_corpus/ 57 0.02 GB fhkh85lzmf
mario_maf_m1_corpus wm-diamond_experiments/mario_maf_m1_corpus/ 10 0.00 GB ks37upxyj8
mario_maf_m1b_human_play wm-diamond_results/mario_maf_m1b_human_play/ 42 0.36 GB fhkh85lzmf
mario_maf_m3_vjepa_level_scaling mario_maf_m3_vjepa_level_scaling/ 132 15.99 GB ks37upxyj8
mario_maf_m3_vjepa_level_scaling wm-diamond_experiments/mario_maf_m3_vjepa_level_scaling/ 16 0.00 GB ks37upxyj8
o0_oracle_ranking wm-diamond_results/o0_oracle_ranking/ 6 0.00 GB uszvmpyjnb
o1_model_ranking o1_model_ranking/ 1 0.04 GB uszvmpyjnb
o1_model_ranking wm-diamond_results/o1_model_ranking/ 7 0.00 GB uszvmpyjnb
o2_causal_split o2_causal_split/ 1 0.00 GB uszvmpyjnb
o2_causal_split wm-diamond_results/o2_causal_split/ 4 0.00 GB uszvmpyjnb
o3_predictor_horizon wm-diamond_results/o3_predictor_horizon/ 4 0.00 GB uszvmpyjnb
o4_temporal_abstraction wm-diamond_results/o4_temporal_abstraction/ 12 0.05 GB uszvmpyjnb
o5_closed_loop_control wm-diamond_results/o5_closed_loop_control/ 3 0.00 GB uszvmpyjnb
o6_dash_jump_diagnostic wm-diamond_results/o6_dash_jump_diagnostic/ 6 0.00 GB uszvmpyjnb
o6_temporal_cem wm-diamond_results/o6_temporal_cem/ 3 0.00 GB uszvmpyjnb
o6e_rollout16 wm-diamond_results/o6e_rollout16/ 16 0.18 GB uszvmpyjnb
o6g_goal_closed_loop wm-diamond_results/o6g_goal_closed_loop/ 3 0.00 GB uszvmpyjnb
o6g_goal_closed_loop_gpu wm-diamond_results/o6g_goal_closed_loop_gpu/ 2 0.00 GB uszvmpyjnb
o6h_skill_commitment wm-diamond_results/o6h_skill_commitment/ 3 0.00 GB uszvmpyjnb
o6i_sticky_cem wm-diamond_results/o6i_sticky_cem/ 3 0.00 GB uszvmpyjnb
o7r_reference_clear wm-diamond_results/o7r_reference_clear/ 3 0.00 GB uszvmpyjnb
o8e_error_escalation wm-diamond_results/o8e_error_escalation/ 4 0.00 GB uszvmpyjnb
o9a_pregate_artifacts wm-diamond_results/o9a_pregate_artifacts/ 6 0.00 GB uszvmpyjnb
o9b2_f2_conditioning o9b2_f2_conditioning/ 5 0.00 GB uszvmpyjnb
o9b2_f2_conditioning wm-diamond_results/o9b2_f2_conditioning/ 15 0.15 GB uszvmpyjnb
o9b_dmacro_f2 wm-diamond_results/o9b_dmacro_f2/ 19 0.28 GB uszvmpyjnb
w1.md … w2.md (DIAMOND w-series; wm-diamond legacy) w8/ 2 0.49 GB uszvmpyjnb
w5 wm-diamond_results/w5/ 1 0.05 GB uszvmpyjnb
w6 wm-diamond_results/w6/ 1 0.00 GB uszvmpyjnb
w6d wm-diamond_results/w6d/ 1 0.00 GB uszvmpyjnb
w8 wm-diamond_results/w8/ 20 0.09 GB uszvmpyjnb

Weight files (final / EMA checkpoints)

path size sha256
mario_maf_m3_vjepa_level_scaling/codec_ema_m2_chunk8.pt 103 MB 922b95a09872cb0f…
mario_maf_m3_vjepa_level_scaling/runs/levels1500/checkpoint_030000.pt 2609 MB e366e94cdb27b269…
mario_maf_m3_vjepa_level_scaling/runs/levels1500/checkpoint_030000_ema.pt 704 MB d5042e8b10dab5d7…
mario_maf_m3_vjepa_level_scaling/runs/levels4000/checkpoint_030000.pt 2609 MB 493aebf69c24b883…
mario_maf_m3_vjepa_level_scaling/runs/levels4000/checkpoint_030000_ema.pt 704 MB f4054d110949c912…
mario_maf_m3_vjepa_level_scaling/runs/levels500/checkpoint_030000.pt 2609 MB 536aec285db7d37f…
mario_maf_m3_vjepa_level_scaling/runs/levels500/checkpoint_030000_ema.pt 704 MB ff17d634324ecd54…
mario_maf_m3_vjepa_level_scaling/runs/levels7201/checkpoint_030000.pt 2609 MB d8bdceaf1ffaea58…
mario_maf_m3_vjepa_level_scaling/runs/levels7201/checkpoint_030000_ema.pt 704 MB f192fb247c841733…
mario_maf_m3_vjepa_level_scaling/runs/levels7201_winonly/checkpoint_030000.pt 2609 MB e7e91294e41ee424…
o1_model_ranking/o1_w6_checkpoints.tgz 43 MB 7742019bbd4fbb5b…
wm-diamond_results/o4_temporal_abstraction/predictor/predictor_best.pt 46 MB fbd22bc00b247dc6…
wm-diamond_results/o6e_rollout16/training/rollout16/predictor_best.pt 46 MB c47f25d7ebecbdac…
wm-diamond_results/o6e_rollout16/training/rollout8/predictor_best.pt 46 MB f502a1965f35813b…
wm-diamond_results/o6e_rollout16/training_retry_interrupted/rollout16/predictor_best.pt 46 MB 1ba66edfd45bef88…
wm-diamond_results/o6e_rollout16/training_retry_interrupted/rollout8/predictor_best.pt 46 MB f502a1965f35813b…
wm-diamond_results/o9b2_f2_conditioning/train/b2_cls_only/predictor_best.pt 47 MB 4a7be3fa61e71965…
wm-diamond_results/o9b2_f2_conditioning/train/b2_full/predictor_best.pt 47 MB fa6ae78d6e9fbc00…
wm-diamond_results/o9b2_f2_conditioning/train/b2_nce_only/predictor_best.pt 47 MB 9da2ffe9c862f071…
wm-diamond_results/o9b_dmacro_f2/pod/train/f2_baseline/predictor_best.pt 46 MB 04b6d7cefc18ea24…
wm-diamond_results/o9b_dmacro_f2/train/f2_baseline/predictor_best.pt 46 MB 04b6d7cefc18ea24…
wm-diamond_results/o9b_dmacro_f2/train/f2_large/predictor_best.pt 131 MB 954dd3c69e19b29e…
wm-diamond_results/o9b_dmacro_f2/train/f2_rollout_only/predictor_best.pt 46 MB ef3573ede18de99b…
w8/w8_persistent_bundle.tgz 489 MB 3d1029db421a2d16…
wm-diamond_results/w5/train/predictor/predictor_best.pt 46 MB 6f772f1dab10cab2…
wm-diamond_results/w8/train/predictor/predictor_best.pt 46 MB d0da10e636c52ad4…
wm-diamond_results/w8/train/probe_split_predictor.pt 46 MB d47ac7f2ff04b290…
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support