Search results

Retokenization drift

Status: Open.

#15 In stanford-cs336/assignment5-alignment;

· phisiart opened on Jun 18, 2026

CUDA and other requirements

Status: Open.

#14 In stanford-cs336/assignment5-alignment;

· rwalle opened on Jun 4, 2026

Missing length normalization for grpo

Status: Open.

#10 In stanford-cs336/assignment5-alignment;

· stepartm opened on Feb 1, 2026

alpaca_sft.prompt problem

Status: Open.

#9 In stanford-cs336/assignment5-alignment;

· chaser682 opened on Dec 5, 2025

In grpo series of experiments, should we fix the learning rate at the very beginning?

Status: Open.

#8 In stanford-cs336/assignment5-alignment;

· heng380 opened on Sep 14, 2025

Is the testcase of dpo loss correct?

Status: Open.

#4 In stanford-cs336/assignment5-alignment;

· tfzzzh opened on Aug 16, 2025

Non-Stanford (following along): Is the Qwen 2.5 Math 1.5B different to HF?

Status: Open.

#3 In stanford-cs336/assignment5-alignment;

· jashvira opened on Aug 9, 2025