Theo Mercer is the editorial pen name used for research and historical synthesis published by the CodeHistory Research Desk. Articles are sourced, reviewed, and revised according to CodeHistory’s published research standards.
SWE-agent reframed language models as software-engineering agents that navigate repositories, edit files and run tests through a specially designed agent-computer interface, moving beyond isolated function generation.
DeepCoder combined neural prediction with classical program search, showing an early way for deep learning to guide synthesis from input-output examples without generating complete source code directly.
Meta AI's TransCoder applied unsupervised machine translation to source code, learning to translate functions among C++, Java and Python without relying on a large parallel corpus of aligned programs.
Salesforce Research's CodeT5 adapted the T5 encoder-decoder architecture to programming languages with identifier-aware objectives, creating one model family for both code understanding and generation tasks.
Meta AI's InCoder trained a generative model to fill missing regions using both left and right context, helping establish infilling as a core capability for code editing rather than simple end-of-file completion.
Salesforce's CodeGen released open code models up to 16.1 billion parameters and explored multi-turn program synthesis, framing coding as a conversation that could decompose a larger intent into successive prompts.
DeepSeek-Coder trained open code models from 1.3B to 33B parameters on a two-trillion-token project-level corpus, combining code-heavy pretraining, long context and infilling during the rapid open-model race of 2023–2024.
Alibaba's Qwen2.5-Coder expanded open code models across six parameter sizes, large multilingual training corpora and long contexts, aiming to combine strong coding ability with preserved math and general-language competence.
LiveCodeBench introduced continuously refreshed programming-contest problems with known publication dates, giving code-model evaluation a way to test on tasks likely to be newer than a model's training cutoff.
Salesforce's CodeRL treated a code language model as an actor and learned critic models from unit-test outcomes, bringing functional execution feedback into both training and inference for program generation.