4-5B Parameter LLMs — Page 104

rt-sam.backdoor_9_lr1e-5_rho0.01

rt-sam.backdoor_9_lr1e-5_rho0.05

rt-sam.backdoor_9_lr1e-5_rho0.1

rt-sam.backdoor_9_lr3e-5_rho0.1

rt-broad_RT.backdoor_9_lr1e-5

rt-broad_RT.backdoor_9_lr3e-5

rt-broad_RT.backdoor_81_lr1e-5

rt-sam.backdoor_81_lr1e-5_rho0.1

Initial-Dual-Reasoning-4B-Added-Special-Tokens

Affine-tsinghuaa2-20251203-085023

k20-lr1e-6-ema0.01-qwen3-4b-think-essay_sensitive50pct-pos_gap50pct

qwen3-4B-instruct-refiner-rl-lr5e-6-step250

gemma3-4b-gsm-sft

z0406_rt_broad_RT_backdoor_0_lr3e-6

z0406_rt_ordinary_RT_backdoor_0_lr1e-6

z0406_rt_broad_RT_backdoor_0_lr1e-5

z0406_rt_broad_RT_backdoor_1_lr1e-6

z0406_rt_broad_RT_backdoor_1_lr1e-5

z0406_rt_ordinary_RT_backdoor_0_lr1e-5

absa-qwen3-4b-instruction-v1

z0406_rt_sam_RT_backdoor_0_lr3e-5_rho0.01

z0406_rt_ordinary_RT_quirk_0_lr1e-4

z0406_rt_sam_RT_backdoor_0_lr3e-5_rho0.02

z0406_rt_sam_RT_backdoor_1_lr3e-5_rho0.01

z0406_rt_ordinary_RT_backdoor_1_lr1e-4

z0406_rt_ordinary_RT_backdoor_1_lr2e-5

z0406_rt_ordinary_RT_backdoor_1_lr5e-5

k10-lr5e-7-ema0.01-qwen3-4b-think-essay_sensitive20pct-pos_gap20pct

qwen3_30b_a3b_to_4b_onpolicy_5k_src20k-25k

router-grpo-v3-merged

vip_grpo_base_p32_2403_qwen3_4b_math__1__1774385112_step1000

reliquary-math

bs16-k10-lr5e-7-ema0.01-eopd0.8-qwen3-4b-think-sciknoweval_chem_bottom20_nogap-maxsteps150

bs16-k10-lr5e-7-ema0.01-eopd0.8-qwen3-4b-think-sciknoweval_material_bottom20_nogap-maxsteps150

hanabi_new_25k_data

intuitor-sciknoweval_material-qwen3-4b-think-2507-r6k100

intuitor-sciknoweval_bio-qwen3-4b-think-2507-r6k100

TTRL-sciknoweval_material-TTRL-Len-8k-grpo-094908

bs16-k20-lr5e-7-ema0-eopd0.8-qwen3-4b-think-mmlu_pro_train10k_bottom20-s150

qwen3-4b-instruct-medium1

math_think_X_qwen3_4b_base_sft

bs16-k10-lr5e-7-ema0.01-eopd0.8-qwen3-4b-think-sciknoweval_material_bottom20_nogap_randret-maxst