Ctrl+K
- 005_logit_lens_probe
- 018_macro_deep
- 020_macro_wide
- 024_trajectory_curriculum
- 025_027_convergence_marathon
- 029_032_causal_interventions
- 033_echo_falsification
- 034_asymmetric_depth
- 035_decoupled_routing
- 036_standard_moe
- 036b_iso_param_cosine
- 036c_bigger_experts
- 037_routing_steering
- 038_expert_suppression
- 039_expert_surgery
- 041_progressive_topk
- 042_multi_seed_wide
- 044_dense_iso_param
- 044c_halting_converged
- 048_bootstrap_ci
- 049_multi_seed_all
- 051_sparsity_ablation
- 053_zipf_frequency_control
- 054_expert_pool_scaling
- 055_active_compute_scaling
- 056_nonlinear_composition
- 057_rank1_sra_mirror
- 058_se_gating
- 059_mixture_of_layers
- 060_sparse_mol
- 061_dense_mol
- 062_attention_modes
- 063_sparse_dispatch
- 065_broader_topology
- 067_openwebtext_replication
- 069_category_discovery
- 070_fullrank_equifinality
- 072_dense_parity
- 073_fair_baselines
- 074_dthin_ablation
- 075_cosmopedia_mol
- 076_moe_ffn_baseline
- 077_cosmopedia_moe_ffn
- 078_cosmopedia_dense_deltanet
- 085_dense_softmax_1p3b
- 087_mol_hybrid_1p3b
- 089_dense_reduced_ffn_1p3b
- 1.52 kB