ViperEkura e149997200 Add DPO data generation & training; fix checkpoint labels; switch to WSD scheduler; update optimizer ablations in abstract/conclusion

2026-07-19 15:25:42 +08:00

..

ckpt_comparison.png

Fix training budget (19k steps / ~20B tokens), add Muon optimizer, correct variance scaling math, move figures to §Training Config, tighten abstract

2026-07-08 22:49:43 +08:00

ckpt_weight_density_per_run.png

Add effective rank analysis, weight distribution evolution, Muon/SFT training figures

2026-07-17 23:05:21 +08:00

ckpt_weight_density.png

Add effective rank analysis, weight distribution evolution, Muon/SFT training figures

2026-07-17 23:05:21 +08:00

dpo_metrics.png

Add DPO training subsection with dpo_metrics figure and Rafailov et al. citation

2026-07-19 15:01:23 +08:00

ifd_length_grid.png

Restructure data pipeline: replace IFD analysis with length bias appendix

2026-07-05 18:59:29 +08:00

loss_compare.png

refactor: numerical stability analysis with residual scaling comparison

2026-06-28 14:36:12 +08:00

muon_pt.png

Add effective rank analysis, weight distribution evolution, Muon/SFT training figures

2026-07-17 23:05:21 +08:00

sft_metrics.png

Update sft_metrics.png

2026-07-18 00:52:52 +08:00