# Alexey Gavrilov - research

Research on discrete latent representations, compression, controllable generation, and inspectable software-artifact regeneration.

Canonical site: https://aogavrilov.com/
Human-readable site index: https://aogavrilov.com/sitemap/
Author profile: https://aogavrilov.com/about/
ORCID: https://orcid.org/0009-0006-3147-5430

## Publications

- [Where Quality Breaks in Compressed Short-Text Generation: Staged Bottleneck Localization](https://aogavrilov.com/publications/where-quality-breaks/): Separating codec failures from latent-generation failures in compressed short-text generation.
- [Inspectable Control for Structure-Preserving Software Regeneration](https://aogavrilov.com/publications/inspectable-control/): Using hierarchical discrete codes as inspectable controls over what a code generator preserves and changes.

## Research guides

- [How to tell whether a compressed text generator fails in the codec or the generator](https://aogavrilov.com/projects/codec-bottleneck-diagnosis/)
- [How can AI edit code without regenerating the entire program?](https://aogavrilov.com/projects/discrete-latent-generation/)

## Focused research notes

- [Code-space vs token-space masked diffusion: how to compare them](https://aogavrilov.com/research-notes/code-space-vs-token-space-masked-diffusion/): A stage-consistent comparison protocol for code-space and token-space masked diffusion language models when the discrete codec is lossy.
- [Localized code modification with generative models](https://aogavrilov.com/research-notes/localized-code-modification-generative-models/): How to prevent unnecessary whole-function rewriting while preserving enough freedom for a generative model to make the requested code change.
- [Constrained code generation for software engineering](https://aogavrilov.com/research-notes/constrained-code-generation-software-engineering/): A practical distinction between grammar constraints, type constraints, preservation boundaries, and behavior-level acceptance checks for generated code.
- [AI-assisted refactoring: methods and evidence](https://aogavrilov.com/research-notes/ai-assisted-refactoring-evidence/): How to evaluate recent AI-assisted refactoring methods without confusing a plausible generated patch with verified behavior preservation.
- [Predictable code generation needs a preservation contract](https://aogavrilov.com/research-notes/predictable-code-generation-preservation-contract/): Why deterministic sampling is not enough, and how observable protected properties and acceptance checks make code-generation behavior testable.

## Machine-readable entry points

- Publications: https://aogavrilov.com/publications.json
- Research-guide router: https://aogavrilov.com/research-guides.json
- Research graph: https://aogavrilov.com/research-graph.jsonld
- LLM index: https://aogavrilov.com/llms.txt
- Full LLM corpus: https://aogavrilov.com/llms-full.txt
- OpenAPI: https://aogavrilov.com/openapi.json
- OAI-PMH: https://aogavrilov.com/oai?verb=Identify