2-3B Parameter LLMs — Page 132

qwen3_1.7b_webshop_macro_action_new_epoch2

teacher_prefix_minesweeper_kukurasu_continual_Qwen3_4B_Thinking_qwen3-1.7b

medical_1bmix_m32-f7a64807-not_easy_1e-4_1200

Qwen3-1.7B-GOPD-DeepMath

math_btoracle-4b-f3c36853-not_easy_1e-4_200

MaxRL-Qwen3-1.7B-Base-IDK-math12k-32-brier-rloo-step2000

Qwen3-1.7B-Wordle-SFT

medmcqa-Qwen2.5-3B-finetuned

Llama-3.2-3B-Instruct_grpo_adv_rollout_8_step580

Llama-3.2-3B-Instruct_base_grpo_rollout_8_20260429_145817_step580

llama-3.2-3b-instruct-only-sn-tuned-lr5e-5

Qwen3-1.7B-Base-dapo_filter-grpo-useKL_True-KLlossCoef1e-3

KG-R1-CWQ-hit1-no-turn-advantage

qwen2.5_sft_merged_dk_it

KG-R1-CWQ-no-retrieval-reward

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_20260429_004543_step580

Qwen3-1.7B-Base-dapo_filter-grpo-noKL

math_m32-4b-9e032637-not_easy_1e-4_800

phi2-docstring-model

llama3_2_3b_instruct_MATH_lr5e-5

filing-sense-grpo-qwen2.5-3b

math_skywork-v2-qwen3-4b-easy_1e-4_200

fake_english_advshape_policyshape_qwen3-1.7b-base

Qwen3-1.7B-Base-dapo_filter-prm-eta100-Advorm-stepsplit-none

KG-R1-CWQ-no-turn-reward

qwen2.5-3b-dora-illnesses

general_knowledge_model

Llama-3.2-3B-Instruct_grpo_adv_rollout_8_20260502_233833_step580

qw3vl2b_evq_grp

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_20260501_120104_step580

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_20260501_115927_step580

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_resume_epoch10_20260429_004543_step232

Llama-3.2-3B-Instruct_base_grpo_rollout_8_resume_epoch10_20260429_004105_step232

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_resume_epoch10_20260429_004543_step290

Llama-3.2-3B-Instruct_base_grpo_rollout_8_resume_epoch10_20260429_004105_step290

group_model

safety_model

group_model

Qwen3-1.7B-icl-3shot-v4_128k-copy_tag-dpo-balanced

math_m32-1b-3d7129ad-not_easy_1e-4_200

math_btoracle-1b-0609ce76-not_easy_1e-4_200

Qwen3-1.7B-MATH-A9-U-TA-GRPO