Small language models (SLMs, ≤1.5B parameters) are attractive for embedded and resource-limited development workflows because they can run under single-GPU or CPU budgets and be adapted without distributed training. SLM-based code generation is brittle under strict sandboxed evaluation, and reinforcement learning (RL) with binary test rewards is often too sparse to train SLMs reliably. This WIP paper presents a reliability-first RL framework for SLM code generation built around a joint reward. The functional term assigns intermediate credit to near-miss outcomes for syntax validity, crash-free execution, and output production, while a static-analysis term discourages unsafe shortcuts during training. On DeepSeek-Coder-1.3B evaluated on 100 stdin-style APPS+ prompts, a binary-to-partial-credit curriculum improves syntax validity to 63% and produces solutions that pass at least one test in 9% of prompts in a single generated attempt. In contrast, binary-reward PPO regresses below a supervised fine-tuning baseline and partial-credit training from scratch reaches only 27% syntax validity.

Tue 16 Jun

Displayed time zone: Mountain Time (US & Canada) change

10:40 - 12:20
Session 3: Formal Methods and Systems ReliabilityLCTES at Flatirons 3
Chair(s): Sanjiva Prasad Indian Institute of Technology Delhi
10:40
22m
Talk
Towards Verifiable System Code using a DSL Compiled to Efficient and Readable C CodeArtifacts AvailableArtifacts Evaluated
LCTES
Clément Chavanon Inria, Univ Rennes, CNRS, IRISA, Henrik Karlsson KTH Royal Institute of Technology, Frédéric Besson Univ Rennes, Inria, CNRS, IRISA, Sandrine Blazy University of Rennes, Roberto Guanciale KTH Royal Institute of Technology
DOI
11:02
22m
Talk
A Pointer-Ownership Model for C Inspired by RustResults ReproducedArtifacts AvailableArtifacts Evaluated
LCTES
David Svoboda , William Klieber Software Engineering Institute, Carnegie Mellon University, Lori Flynn CERT, Ruben Martins Carnegie Mellon University, Jeffrey Hoskinson Software Engineering Institute, Carnegie Mellon University
DOI
11:24
22m
Talk
Hikami: A Lightweight Hypervisor for Emulating RISC-V Extension Semantics with Sail-Driven Auto-generationResults ReproducedArtifacts Available
LCTES
Norimasa Takana University of Tsukuba, Yoshihiro Oyama University of Tsukuba
DOI
11:46
10m
Short-paper
Scheduled Partial-Credit RL for Reliable Code Generation with Small Language Models (WIP)
LCTES
Suryansh Singh Sijwali The Pennsylvania State University, Suman Saha pc
DOI