Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models

• 类型：arxiv
• 标识：2607.12463
• 链接：http://arxiv.org/abs/2607.12463v1
• 主分类：agent
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar + OpenAlex
• S2被引：0
• OpenAlex被引：0
• 影响力被引：0
• TLDR：Beyond in-domain gains, mid-training mitigates the capability erosion that agentic post-training otherwise inflicts on non-agent coding and non-coding tool-use benchmarks (tau-bench, BFCL): although the mid-training corpus contains Python code only, the function-call inductive bias survives post-training and yields consistent gains.
• OpenAlex ID：W7168372324
• OpenAlex DOI：10.48550/arxiv.2607.12463
• DOI：10.48550/arxiv.2607.12463
• DOI来源：OpenAlex
• 开放获取：green
• 开放获取链接：https://doi.org/10.48550/arxiv.2607.12463
• OpenAlex更新：2026-07-19
• 副分类：engineering
• 待LLM分类：否
• 标题中文：面向 Coding Agent 基础模型的 Function-Aware Fill-in-the-Middle 中期训练
• TLDR中文：除领域内增益外，mid-training 还能缓解 agentic post-training 对非 Agent 编程及非编程工具调用基准（tau-bench、BFCL）造成的能力侵蚀：尽管 mid-training 语料仅含 Python 代码，函数调用的归纳偏置在 post-training 后依然保留，带来稳定的增益。
• 来源文件：
• /inbox/tom/_candidates/2026-07-16-agent-rag-longcontext-candidates.json
• [S2 enrich]
• [OpenAlex backfill]