Skip to content

JayZenith/GLYPH

Repository files navigation

GLYPH

Post-training a 4B Rust coding agent with verified synthetic traces, full SFT, verifier RL, pass@8 evaluation, and a live TUI.

ASSISTANT  CALL read_file {...}
TOOL       RESULT src/lib.rs
ASSISTANT  CALL apply_patch {...}
ASSISTANT  CALL cargo_test {...}
TOOL       RESULT 4 passed
ASSISTANT  FINAL: fixed and tested

The full conversation is the training unit. Real Cargo execution verifies the data and rewards. verifiers owns the environment; PRIME-RL runs RL.

Write-up · Environment · SFT model · Dense RLVR · Evaluation files

Results

valid@8: at least one of eight rollouts passes Cargo and ends with one clean FINAL.

Model valid@8 / 150, retained runs
SFT 95
Sparse RLVR 98 / 96 / 98
Dense RLVR 102
Compiler-aware RLVR 95 / 96 / 94

No retained-run comparison reached p < 0.05. Methodology and interpretation

Run the environment

Requires Python 3.10+, Cargo, and rustc.

git clone https://github.com/JayZenith/GLYPH.git
cd GLYPH
python3 -m venv .venv
source .venv/bin/activate
pip install -e environments/glyph
vf-eval glyph -m <your-model> -n 20 -r 5

Interactive TUI smoke test

The TUI shows the exact ChatML system, user, assistant, and tool turns while the model patches a disposable Rust crate.

Run the live vLLM demo or offline preview

Safety

Model-edited Rust is arbitrary code. Use a disposable VM, container, or isolated job. Host Cargo execution is the default; Bubblewrap is opt-in.

Verify or reproduce

Source layout · Companion Rust crates

About

post-training stack for a 4B Rust coding agent: synthetic data, SFT, verifier RL, and pass@8 evaluation.

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages