-
oaimli/pgpo_grpo_full_scitrek_qwen3_4b_instruct_2507
Text Generation • 4B • Updated • 159 -
oaimli/pgpo_grpo_full_scitrek_gemma3_4b_it
Image-Text-to-Text • 4B • Updated • 34 -
oaimli/pgpo_grpo_proxy_scitrek_qwen3_4b_instruct_2507
Text Generation • 4B • Updated • 109 -
oaimli/pgpo_grpo_proxy_scitrek_gemma3_4b_it
Image-Text-to-Text • 4B • Updated • 95
Miao Li
oaimli
AI & ML interests
Natural Language Processing
Recent Activity
updated a model 3 days ago
oaimli/sft_longpt_full_qwen3_4b_instruct_2507 published a model 3 days ago
oaimli/sft_longpt_full_qwen3_4b_instruct_2507 updated a model 4 days ago
oaimli/sft_longpt_full_gemma3_4b_itOrganizations
None yet
Models for Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning (ACL 2026)
-
oaimli/longtune_scitrek_reasoning_reinforcement_qwen
Text Generation • 4B • Updated • 10 -
oaimli/longtune_scitrek_grounding_reinforcement_qwen_5_300
Text Generation • 4B • Updated • 10 -
oaimli/longtune_scitrek_grounding_reinforcement_qwen_0_300
Text Generation • 4B • Updated • 6 -
oaimli/longtune_scitrek_reasoning_reinforcement_gemma
Image-Text-to-Text • 4B • Updated • 10
Long-context post-training
Models for the paper of "Who Gets Cited Most? Benchmarking Long-Context Language Models on Scientific Articles"
-
oaimli/scitrek_qwen25_7b_instruct_1m_sft_input_length
Text Generation • 8B • Updated • 15 -
oaimli/scitrek_qwen25_7b_instruct_1m_sft_question_focus
Text Generation • 8B • Updated • 5 -
oaimli/scitrek_qwen25_7b_instruct_1m_sft_sql_type
Text Generation • 8B • Updated • 8 -
oaimli/scitrek_qwen25_7b_instruct_1m_grpo_curriculum_192
Text Generation • 8B • Updated • 30
-
oaimli/pgpo_grpo_full_scitrek_qwen3_4b_instruct_2507
Text Generation • 4B • Updated • 159 -
oaimli/pgpo_grpo_full_scitrek_gemma3_4b_it
Image-Text-to-Text • 4B • Updated • 34 -
oaimli/pgpo_grpo_proxy_scitrek_qwen3_4b_instruct_2507
Text Generation • 4B • Updated • 109 -
oaimli/pgpo_grpo_proxy_scitrek_gemma3_4b_it
Image-Text-to-Text • 4B • Updated • 95
Long-context post-training
Models for Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning (ACL 2026)
-
oaimli/longtune_scitrek_reasoning_reinforcement_qwen
Text Generation • 4B • Updated • 10 -
oaimli/longtune_scitrek_grounding_reinforcement_qwen_5_300
Text Generation • 4B • Updated • 10 -
oaimli/longtune_scitrek_grounding_reinforcement_qwen_0_300
Text Generation • 4B • Updated • 6 -
oaimli/longtune_scitrek_reasoning_reinforcement_gemma
Image-Text-to-Text • 4B • Updated • 10
Models for the paper of "Who Gets Cited Most? Benchmarking Long-Context Language Models on Scientific Articles"
-
oaimli/scitrek_qwen25_7b_instruct_1m_sft_input_length
Text Generation • 8B • Updated • 15 -
oaimli/scitrek_qwen25_7b_instruct_1m_sft_question_focus
Text Generation • 8B • Updated • 5 -
oaimli/scitrek_qwen25_7b_instruct_1m_sft_sql_type
Text Generation • 8B • Updated • 8 -
oaimli/scitrek_qwen25_7b_instruct_1m_grpo_curriculum_192
Text Generation • 8B • Updated • 30