2-3B Parameter LLMs — Page 49

qwen3_1.7b_easy_rl_reinforce_alpha_0

qwen3_1.7b_easy_rl_final_step120

qwen3_1.7b_easy_rl_ours_adv_fixed_gamma_995_98_ori_norm

Vanilla_RL

64b_SFT

qwen3_1.7b_easy_rl_ours_adv_fixed_gamma_1_98_geo_ms_token_tis

qwen3_1.7b_easy_rl_ours_adv_fixed_geo_ms_token_tis

qwen3_1.7b_new_standard_A_sft_overfit_lr_5e_6__global_step_192

qwen3_1.7b_easy_rl_ours_adv_fixed_no_norm

qwen3_1.7b_new_standard_A_sft_overfit_lr_5e_6__global_step_288

qwen3_1.7b_new_standard_C_sft_overfit_lr_5e_6__global_step_296

2b_RL_DAPO

2b_SFT

qwen3_1.7b_new_sudoku_one_action_A_sft_lr_5e_6__step_1124

qwen3_1.7b_new_sudoku_one_action_B_sft_lr_5e_6__step_1108

north_llama32_3b_enhancedNCC_instruct_v1_long_large_lr2e6_2048_90000

qwen3_1.7b_rush_hour_multi_move_final_new

qwen3-1.7b-base-adam-1e-6-bs128-kl0.0-global_step_200

evolved_set1_correct_12k_ep10

qwen3-1.7b-base-combined-sft-ckpt-360

99-caldpo-dataset-our-39-llama3-2-3b-instruct-merged

qwen3_1.7b_one_act_easy_short

qwen3-1.7b-base-svd-muon-adam-1e-6-bs128-kl0.0-global_step_20

qwen3_1.7b_sudoku_one_action_easy_21_30_epoch1

qwen3_1.7b_rush_hour_multi_move_final_4_9_long_10_12_epoch3

Qwen3-1.7B-CCC-merged-cp2-LR1e-4

Qwen3-1.7B-CCC-merged-cp3-LR1e-4

Qwen3-1.7B-CCC-merged-cp5-LR1e-4

dl_finetuned_minicoder

qwen3-1.7b-amr-20260204-1017

LogicBench-Qwen-FT-Response

qwen2.5-3B-distill-Math-Alpaca

Qwen3-1.7B-Tiny-Hanabi-XML-SFT-4

DAPO_1.7B_step120

ghost-engine-v2-merged

Qwen3-1.7B-Tiny-Hanabi-XML-SFT-5

llama-32-3b-instruct-openthoughts-nothink-8192-epoch1.0-bs4

llama-32-3b-midtrain-openthoughts-nothink-8192-epoch3.0-bs4

unified-model-stage1-action-tokens-v2

llama-32-3b-midtrain-openthoughts-8192-epoch3.0-bs4

qwen3-1.7b-amr-20260204-1342

dexter-merged