2-3B Parameter LLMs — Page 129

fintech_gemma_2b

gemma_2b_it_fintech

qwen25-3b-somali

Qwen3-1.7B-Wanda_4_8

fintech_gemma_2b

Qwen3-1.7B-Wanda_1_4

Qwen3-1.7B-Wanda_unstruct_0.5

AutoGraphR1-musique_hotpotqa_train-llama3.2-3b-text-retriever-grpo-repetition-penalty

qwen2.5_sft_merged_dk_it

KG-R1-CQW

Qwen3-VL-2B-Instruct-Docling-5K-30perc-11ep

stack-x-ultimate-v2

markovify_advshape_policy_shape_qwen3-1.7b-base

science_skywork_reward_v2_qwen3_4b_not_easy_1e-4_400

Qwen2.5-Coder-3B-SFT-WebCode

qwen2.5_3b_instruct_finetuned

llama3_2_3b_instruct_only_rsn_tuned_lr5e-5

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

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_Use_KL_0.001_step580

evolai-qwen3-1.7b-v1

qwen2.5-3b-dora-illnesses

qw3vl2b_ifs_grp

Llama-3.2-3B-Instruct_grpo_adv_rollout_8_20260502_233833_step580

CASLIE-S

qwen3-1.7b-sft-3

group_model

Qwen3-1.7B-Wordle-SFT

safety_model

Llama-3.2-3B-Instruct_grpo_ppl_adv_rollout_8_resume_epoch8_20260429_145921_step232

qw3vl2b_ifs

safety_model

math_model

my-style-model

grpo-3b

Qwen2.5-VL-3B-Instruct-ft_lang

SearchR1-nq_hotpotqa_train-qwen2.5-3b-em-grpo-v0.2

qwen3_1.7b_clipcov_verified_grpo_eq3ep

qwen3_1.7b_klcov_verified_grpo

qwen3_1.7b_baseline_full_grpo

qwen3_1.7b_clipcov_verified_grpo

qwen3_1.7b_baseline_verified_grpo

qwen3_1.7b_vdrop75_verified_grpo