# Synthetic Hallucinations, Real Gains: Hard Negatives from Frontier Models for FIM Hallucination Mitigation

arXiv: https://arxiv.org/abs/2606.03130 · HTML: https://mlsanity.com/paper/2606.03130/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 15 of its 28 references that are in the map.
- First posted on arXiv 2 Jun 2026.

## Builds on (most cited first, 15 of 15)

- [Self-Instruct: Aligning Language Models with Self-Generated Instructions](https://mlsanity.com/paper/2212.10560/) (2022)
- [LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://mlsanity.com/paper/2403.13372/) (2024)
- [A Survey on LLM-as-a-Judge](https://mlsanity.com/paper/2411.15594/) (2024)
- [DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence](https://mlsanity.com/paper/2401.14196/) (2024)
- [ORPO: Monolithic Preference Optimization without Reference Model](https://mlsanity.com/paper/2403.07691/) (2024)
- [StarCoder 2 and The Stack v2: The Next Generation](https://mlsanity.com/paper/2402.19173/) (2024)
- [InCoder: A Generative Model for Code Infilling and Synthesis](https://mlsanity.com/paper/2204.05999/) (2022)
- [Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models](https://mlsanity.com/paper/2404.18796/) (2024)
- [Efficient Training of Language Models to Fill in the Middle](https://mlsanity.com/paper/2207.14255/) (2022)
- [Self-Edit: Fault-Aware Code Editor for Code Generation](https://mlsanity.com/paper/2305.04087/) (2023)
- [Execution-based Code Generation using Deep Reinforcement Learning](https://mlsanity.com/paper/2301.13816/) (2023)
- [The RealHumanEval: Evaluating Large Language Models' Abilities to Support Programmers](https://mlsanity.com/paper/2404.02806/) (2024)
- [Evaluation of LLMs on Syntax-Aware Code Fill-in-the-Middle Tasks](https://mlsanity.com/paper/2403.04814/) (2024)
- [Structure-Aware Fill-in-the-Middle Pretraining for Code](https://mlsanity.com/paper/2506.00204/) (2025)
- [Delulu: A Verified Multi-Lingual Benchmark for Code Hallucination Detection in Fill-in-the-Middle Tasks](https://mlsanity.com/paper/2605.07024/) (2026)

## Abstract (arXiv)

Small open-source code models that power IDE autocomplete still emit hallucinated Fill-in-the-Middle (FIM) completions: syntactically natural calls to methods, parameters, variables, and imports that do not exist in the surrounding project. Existing mitigations either require per-language execution sandboxes that do not apply at mid-keystroke or preference-optimisation pipelines that need large human-labelled corpora. We propose an execution-free alternative: use frontier code models to synthesise plausible-but-wrong completions as hard negatives, then leverage the contrast between these synthetic hallucinations and the ground-truth developer edit as a supervised fine-tuning signal. Our pipeline scrapes multilingual FIM contexts from public GitHub across eight languages and asks a panel of three frontier generators to produce one hard negative per context for each of four hallucination types drawn from the Delulu taxonomy, a Docker-verified multilingual FIM hallucination benchmark, yielding a paired chosen/rejected dataset. Fine-tuning Qwen2.5-Coder-7B-Instruct on a 100K-row curated subset lifts Delulu exact match by +18.8 points and edit similarity by +0.22 on every language and every type, while also improving every HumanEval-Infilling split and every SAFIM subset. The same recipe at 3B lifts Delulu by +12.8 EM with a small, characterised general-FIM trade-off. Five-axis ablations (size, type mix, language coverage, base-model family, and a difficulty-aware fool rate) plus a head-to-head SFT vs. DPO/ORPO comparison map which design choices drive the gain. We release the full pipeline source code -- generation, fool-rate LLM judging, curation, and the FIM fine-tuning recipe -- so that the experiments in this paper can be reproduced end-to end on any permissively licensed corpus.