Copertina

State-of-the-art Small Language Coder Model: Mify-Coder

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Parmar, Abhinav, Panigrahi, Abhisek, Dwivedi, Abhishek Kumar, Bhattacharya, Abhishek, Ramachandra, Adarsh, Choudhary, Aditya, Garg, Aditya, Raj, Aditya, Bhatt, Alankrit, Yadav, Alpesh, Vishnu, Anant, Pillai, Ananthu, Kumar, Ankush, Patnaik, Aryan, S, Aswatha Narayanan, Singh, Avanish Raj, Gadda, Bhavya Shree, Kachhadiya, Brijesh Pankajbhai, Jahnavi, Buggala, Krishna, Chidurala Nithin, Shah, Chintan, Akshaya, Chunduru, Banerjee, Debarshi, Dey, Debrup, R., Deepa, G, Deepika B, Rahman, Faiz ur, Gayari, Gagan, Naidu, Gudhi Jagadeesh Kumar, Singh, Gursimar, Tyagi, Harshal, K, Harshini, Vathalloor, James Mani, Nettar, Jayarama, Gajjam, Jayashree, George, Joe Walter Sugil, Tadepalli, Kamalakara Sri Krishna, Rathinasamy, Kamalkumar, Chaurasia, Karan, S, Karthikeyan, Arora, Kashish, Desai, Kaushal, Buwade, Khushboo, Manjrekar, Kiran, Likhitha, Malikireddy Venkata Sai, A, Manjunath, Bedmutha, Mitali Mahavir, Tarafdar, Mohammed Rafee, Tiwari, Nikhil, Gigi, Nikitha K, Ravikumar, Pavan, Swarnanjali, Pendyala, Anand, Piyush, Chandrasekar, Prakash, Gawade, Prasanna Bhalchandra, Sivan, Prasanth, Khurana, Preeti, Babbar, Priyanshi, Mondal, Rajab Ali, Vissapragada, Rajesh Kumar, Ganesan, Rajeshwari, Koppisetti, Rajeswari, R., Ramjee, Thiruppathisamy, Ramkumar, S., Rani G., Reka, S, Gupta, Samarth, Kothakota, Sandeep Reddy, K, Sarathy, Kumar, Sathyanarayana Sampath, Kumar, Saurabh, Khasare, Shashank, Kumar, Shenbaga Devi Venkatesh, Parvatham, Shiva Rama Krishna, Shaikh, Shoeb, A, Shrishanmathi, Pathak, Shubham, Koppaka, Sree Samhita, S, Sreenivasa Raghavan K, Venkatasubramanian, Sreeram, Bojja, Suprabha Desai, R, Swetha, Ahmed, Syed, Thota, Chinmai Harshitha, Yadav, Tushar, Kusumitha, Veeravelly, Patnaik, V V S S Prasanth, Sesetti, Vidya Sri, K, Vijayakeerthi, Bakshi, Vikram Raj, K, Vinay K, Loganathan, Vinoth Kumar, Tiwari, Vipin, Shrivastav, Vivek Kumar, Charan, V Venkata Sri Datta, Khan, Wasim Akhtar
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Software Engineering Artificial Intelligence Computation and Language
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

_version_	1866914224678109184
author	Parmar, Abhinav Panigrahi, Abhisek Dwivedi, Abhishek Kumar Bhattacharya, Abhishek Ramachandra, Adarsh Choudhary, Aditya Garg, Aditya Raj, Aditya Bhatt, Alankrit Yadav, Alpesh Vishnu, Anant Pillai, Ananthu Kumar, Ankush Patnaik, Aryan S, Aswatha Narayanan Singh, Avanish Raj Gadda, Bhavya Shree Kachhadiya, Brijesh Pankajbhai Jahnavi, Buggala Krishna, Chidurala Nithin Shah, Chintan Akshaya, Chunduru Banerjee, Debarshi Dey, Debrup R., Deepa G, Deepika B Rahman, Faiz ur Gayari, Gagan Naidu, Gudhi Jagadeesh Kumar Singh, Gursimar Tyagi, Harshal K, Harshini Vathalloor, James Mani Nettar, Jayarama Gajjam, Jayashree George, Joe Walter Sugil Tadepalli, Kamalakara Sri Krishna Rathinasamy, Kamalkumar Chaurasia, Karan S, Karthikeyan Arora, Kashish Desai, Kaushal Buwade, Khushboo Manjrekar, Kiran Likhitha, Malikireddy Venkata Sai A, Manjunath Bedmutha, Mitali Mahavir Tarafdar, Mohammed Rafee Tiwari, Nikhil Gigi, Nikitha K Ravikumar, Pavan Swarnanjali, Pendyala Anand, Piyush Chandrasekar, Prakash Gawade, Prasanna Bhalchandra Sivan, Prasanth Khurana, Preeti Babbar, Priyanshi Mondal, Rajab Ali Vissapragada, Rajesh Kumar Ganesan, Rajeshwari Koppisetti, Rajeswari R., Ramjee Thiruppathisamy, Ramkumar S., Rani G. Reka, S Gupta, Samarth Kothakota, Sandeep Reddy K, Sarathy Kumar, Sathyanarayana Sampath Kumar, Saurabh Khasare, Shashank Kumar, Shenbaga Devi Venkatesh Parvatham, Shiva Rama Krishna Shaikh, Shoeb A, Shrishanmathi Pathak, Shubham Koppaka, Sree Samhita S, Sreenivasa Raghavan K Venkatasubramanian, Sreeram Bojja, Suprabha Desai R, Swetha Ahmed, Syed Thota, Chinmai Harshitha Yadav, Tushar Kusumitha, Veeravelly Patnaik, V V S S Prasanth Sesetti, Vidya Sri K, Vijayakeerthi Bakshi, Vikram Raj K, Vinay K Loganathan, Vinoth Kumar Tiwari, Vipin Shrivastav, Vivek Kumar Charan, V Venkata Sri Datta Khan, Wasim Akhtar
author_facet	Parmar, Abhinav Panigrahi, Abhisek Dwivedi, Abhishek Kumar Bhattacharya, Abhishek Ramachandra, Adarsh Choudhary, Aditya Garg, Aditya Raj, Aditya Bhatt, Alankrit Yadav, Alpesh Vishnu, Anant Pillai, Ananthu Kumar, Ankush Patnaik, Aryan S, Aswatha Narayanan Singh, Avanish Raj Gadda, Bhavya Shree Kachhadiya, Brijesh Pankajbhai Jahnavi, Buggala Krishna, Chidurala Nithin Shah, Chintan Akshaya, Chunduru Banerjee, Debarshi Dey, Debrup R., Deepa G, Deepika B Rahman, Faiz ur Gayari, Gagan Naidu, Gudhi Jagadeesh Kumar Singh, Gursimar Tyagi, Harshal K, Harshini Vathalloor, James Mani Nettar, Jayarama Gajjam, Jayashree George, Joe Walter Sugil Tadepalli, Kamalakara Sri Krishna Rathinasamy, Kamalkumar Chaurasia, Karan S, Karthikeyan Arora, Kashish Desai, Kaushal Buwade, Khushboo Manjrekar, Kiran Likhitha, Malikireddy Venkata Sai A, Manjunath Bedmutha, Mitali Mahavir Tarafdar, Mohammed Rafee Tiwari, Nikhil Gigi, Nikitha K Ravikumar, Pavan Swarnanjali, Pendyala Anand, Piyush Chandrasekar, Prakash Gawade, Prasanna Bhalchandra Sivan, Prasanth Khurana, Preeti Babbar, Priyanshi Mondal, Rajab Ali Vissapragada, Rajesh Kumar Ganesan, Rajeshwari Koppisetti, Rajeswari R., Ramjee Thiruppathisamy, Ramkumar S., Rani G. Reka, S Gupta, Samarth Kothakota, Sandeep Reddy K, Sarathy Kumar, Sathyanarayana Sampath Kumar, Saurabh Khasare, Shashank Kumar, Shenbaga Devi Venkatesh Parvatham, Shiva Rama Krishna Shaikh, Shoeb A, Shrishanmathi Pathak, Shubham Koppaka, Sree Samhita S, Sreenivasa Raghavan K Venkatasubramanian, Sreeram Bojja, Suprabha Desai R, Swetha Ahmed, Syed Thota, Chinmai Harshitha Yadav, Tushar Kusumitha, Veeravelly Patnaik, V V S S Prasanth Sesetti, Vidya Sri K, Vijayakeerthi Bakshi, Vikram Raj K, Vinay K Loganathan, Vinoth Kumar Tiwari, Vipin Shrivastav, Vivek Kumar Charan, V Venkata Sri Datta Khan, Wasim Akhtar
contents	We present Mify-Coder, a 2.5B-parameter code model trained on 4.2T tokens using a compute-optimal strategy built on the Mify-2.5B foundation model. Mify-Coder achieves comparable accuracy and safety while significantly outperforming much larger baseline models on standard coding and function-calling benchmarks, demonstrating that compact models can match frontier-grade models in code generation and agent-driven workflows. Our training pipeline combines high-quality curated sources with synthetic data generated through agentically designed prompts, refined iteratively using enterprise-grade evaluation datasets. LLM-based quality filtering further enhances data density, enabling frugal yet effective training. Through disciplined exploration of CPT-SFT objectives, data mixtures, and sampling dynamics, we deliver frontier-grade code intelligence within a single continuous training trajectory. Empirical evidence shows that principled data and compute discipline allow smaller models to achieve competitive accuracy, efficiency, and safety compliance. Quantized variants of Mify-Coder enable deployment on standard desktop environments without requiring specialized hardware.
format	Preprint
id	arxiv_https___arxiv_org_abs_2512_23747
institution	arXiv
publishDate	2025
record_format	arxiv
spellingShingle	State-of-the-art Small Language Coder Model: Mify-Coder Parmar, Abhinav Panigrahi, Abhisek Dwivedi, Abhishek Kumar Bhattacharya, Abhishek Ramachandra, Adarsh Choudhary, Aditya Garg, Aditya Raj, Aditya Bhatt, Alankrit Yadav, Alpesh Vishnu, Anant Pillai, Ananthu Kumar, Ankush Patnaik, Aryan S, Aswatha Narayanan Singh, Avanish Raj Gadda, Bhavya Shree Kachhadiya, Brijesh Pankajbhai Jahnavi, Buggala Krishna, Chidurala Nithin Shah, Chintan Akshaya, Chunduru Banerjee, Debarshi Dey, Debrup R., Deepa G, Deepika B Rahman, Faiz ur Gayari, Gagan Naidu, Gudhi Jagadeesh Kumar Singh, Gursimar Tyagi, Harshal K, Harshini Vathalloor, James Mani Nettar, Jayarama Gajjam, Jayashree George, Joe Walter Sugil Tadepalli, Kamalakara Sri Krishna Rathinasamy, Kamalkumar Chaurasia, Karan S, Karthikeyan Arora, Kashish Desai, Kaushal Buwade, Khushboo Manjrekar, Kiran Likhitha, Malikireddy Venkata Sai A, Manjunath Bedmutha, Mitali Mahavir Tarafdar, Mohammed Rafee Tiwari, Nikhil Gigi, Nikitha K Ravikumar, Pavan Swarnanjali, Pendyala Anand, Piyush Chandrasekar, Prakash Gawade, Prasanna Bhalchandra Sivan, Prasanth Khurana, Preeti Babbar, Priyanshi Mondal, Rajab Ali Vissapragada, Rajesh Kumar Ganesan, Rajeshwari Koppisetti, Rajeswari R., Ramjee Thiruppathisamy, Ramkumar S., Rani G. Reka, S Gupta, Samarth Kothakota, Sandeep Reddy K, Sarathy Kumar, Sathyanarayana Sampath Kumar, Saurabh Khasare, Shashank Kumar, Shenbaga Devi Venkatesh Parvatham, Shiva Rama Krishna Shaikh, Shoeb A, Shrishanmathi Pathak, Shubham Koppaka, Sree Samhita S, Sreenivasa Raghavan K Venkatasubramanian, Sreeram Bojja, Suprabha Desai R, Swetha Ahmed, Syed Thota, Chinmai Harshitha Yadav, Tushar Kusumitha, Veeravelly Patnaik, V V S S Prasanth Sesetti, Vidya Sri K, Vijayakeerthi Bakshi, Vikram Raj K, Vinay K Loganathan, Vinoth Kumar Tiwari, Vipin Shrivastav, Vivek Kumar Charan, V Venkata Sri Datta Khan, Wasim Akhtar Software Engineering Artificial Intelligence Computation and Language We present Mify-Coder, a 2.5B-parameter code model trained on 4.2T tokens using a compute-optimal strategy built on the Mify-2.5B foundation model. Mify-Coder achieves comparable accuracy and safety while significantly outperforming much larger baseline models on standard coding and function-calling benchmarks, demonstrating that compact models can match frontier-grade models in code generation and agent-driven workflows. Our training pipeline combines high-quality curated sources with synthetic data generated through agentically designed prompts, refined iteratively using enterprise-grade evaluation datasets. LLM-based quality filtering further enhances data density, enabling frugal yet effective training. Through disciplined exploration of CPT-SFT objectives, data mixtures, and sampling dynamics, we deliver frontier-grade code intelligence within a single continuous training trajectory. Empirical evidence shows that principled data and compute discipline allow smaller models to achieve competitive accuracy, efficiency, and safety compliance. Quantized variants of Mify-Coder enable deployment on standard desktop environments without requiring specialized hardware.
title	State-of-the-art Small Language Coder Model: Mify-Coder
topic	Software Engineering Artificial Intelligence Computation and Language
url	https://arxiv.org/abs/2512.23747