Shubhashis Roy Dipta PRO
dipta007
AI & ML interests
Multimodal Understanding, Reasoning, Generation
Recent Activity
authored a paper 12 days ago
CulturalRiddles: A Multicultural Benchmark of RiddlesOrganizations
-
OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolution
Paper • 2609.06490 • Published • 7 -
dipta007/OracleZoom
Image-to-Image • Updated • 5 -
dipta007/OracleZoom-4KLSDB-train
Viewer • Updated • 52.8k • 432 • 1 -
OracleZoom
🔎14Zoom any photo to 256x, one 4x step at a time
-
DecomposeRL: Learning to Ask Useful, Informative, and Diverse Questions for Semi-Supervised, Traceable Claim Verification
Paper • 2605.27858 • Published -
dipta007/DecomposeRL
Viewer • Updated • 237k • 466 -
dipta007/decomposeRL-7b
Text Generation • 8B • Updated • 25 -
dipta007/decomposeRL-tiny-judge
Viewer • Updated • 20.2M • 776
-
Advancing Reference-free Evaluation of Video Captions with Factual Analysis
Paper • 2509.16538 • Published • 2 -
dipta007/VCInspector-7B
Image-Text-to-Text • 8B • Updated • 79 • 1 -
dipta007/VCInspector-3B
Image-Text-to-Text • 4B • Updated • 74 • 1 -
dipta007/ActivityNet-FG-It
Viewer • Updated • 242k • 349
Synthetically rephrased version of BIRD dataset
A graph based CoT for math reasoning (DAGGER tokens <<<< CoT Tokens)
-
GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO
Paper • 2601.06767 • Published • 1 -
dipta007/Ganit
Viewer • Updated • 32.3k • 57 -
dipta007/GanitLLM-4B_SFT_CGRPO
Text Generation • 196k • Updated • 32 -
dipta007/GanitLLM-4B_SFT_GRPO
Text Generation • 196k • Updated • 30 • 1
Datasets used in the paper: Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
-
Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
Paper • 2506.10202 • Published -
dipta007/Q2E_MultiVENT_LLAMA_3.3_70B_InternVL_38B_Funiform_16_noASR
Viewer • Updated • 2.39k • 6 -
dipta007/Q2E_MultiVENT_LLAMA_3.3_70B_InternVL_38B_Funiform_16_ASR
Viewer • Updated • 2.39k • 16 -
dipta007/Q2E_MSRVTT-1kA_LLAMA_3.3_70B_InternVL_38B_Funiform_16_ASR
Viewer • Updated • 1k • 10
-
OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolution
Paper • 2609.06490 • Published • 7 -
dipta007/OracleZoom
Image-to-Image • Updated • 5 -
dipta007/OracleZoom-4KLSDB-train
Viewer • Updated • 52.8k • 432 • 1 -
OracleZoom
🔎14Zoom any photo to 256x, one 4x step at a time
A graph based CoT for math reasoning (DAGGER tokens <<<< CoT Tokens)
-
DecomposeRL: Learning to Ask Useful, Informative, and Diverse Questions for Semi-Supervised, Traceable Claim Verification
Paper • 2605.27858 • Published -
dipta007/DecomposeRL
Viewer • Updated • 237k • 466 -
dipta007/decomposeRL-7b
Text Generation • 8B • Updated • 25 -
dipta007/decomposeRL-tiny-judge
Viewer • Updated • 20.2M • 776
-
Advancing Reference-free Evaluation of Video Captions with Factual Analysis
Paper • 2509.16538 • Published • 2 -
dipta007/VCInspector-7B
Image-Text-to-Text • 8B • Updated • 79 • 1 -
dipta007/VCInspector-3B
Image-Text-to-Text • 4B • Updated • 74 • 1 -
dipta007/ActivityNet-FG-It
Viewer • Updated • 242k • 349
-
GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO
Paper • 2601.06767 • Published • 1 -
dipta007/Ganit
Viewer • Updated • 32.3k • 57 -
dipta007/GanitLLM-4B_SFT_CGRPO
Text Generation • 196k • Updated • 32 -
dipta007/GanitLLM-4B_SFT_GRPO
Text Generation • 196k • Updated • 30 • 1
Synthetically rephrased version of BIRD dataset
Datasets used in the paper: Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
-
Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
Paper • 2506.10202 • Published -
dipta007/Q2E_MultiVENT_LLAMA_3.3_70B_InternVL_38B_Funiform_16_noASR
Viewer • Updated • 2.39k • 6 -
dipta007/Q2E_MultiVENT_LLAMA_3.3_70B_InternVL_38B_Funiform_16_ASR
Viewer • Updated • 2.39k • 16 -
dipta007/Q2E_MSRVTT-1kA_LLAMA_3.3_70B_InternVL_38B_Funiform_16_ASR
Viewer • Updated • 1k • 10