Preference learning extends the performance of Code LLMs beyond traditional supervised fine-tuning by leveraging relative quality comparisons. In existing approaches, a set of n candidate solutions is evaluated based on test case success rates, with the candidate demonstrating a higher pass rate being labeled as positive and its counterpart with a lower pass rate as negative. However, because this approach aligns entire failing code blocks rather than pinpointing specific errors, it lacks the granularity necessary to capture meaningful error-correction relationships. As a result, the model is unable to learn more informative error-correction patterns. To address these issues, we propose Target-DPO, a new preference alignment framework that mimics human iterative debugging to refine Code LLMs. Target-DPO explicitly locates error regions and aligns the corresponding tokens via a tailored DPO algorithm. To facilitate it, we introduce the CodeFlow dataset, where samples are iteratively refined until passing tests, with modifications capturing error corrections. Extensive experiments show that a diverse suite of Code LLMs equipped with Target-DPO achieves significant performance gains in code generation and improves on challenging tasks like BigCodeBench. In-depth analysis reveals that Target-DPO yields fewer errors. Code, model and datasets are in: https://github.com/JieWu02/Target-DPO.

Event argument extraction aims to identify event arguments and classify their roles within events, whereas relation extraction classifies semantic relationships between entities. Existing methods typically design task-specific models for EAE, which restricts the integration of relation-level semantics. Consequently, they overlook the complementary cues from RE that are beneficial for argument role disambiguation. To overcome this limitation, we propose REAR, a Relation-aware EAE Reinforced optimization framework. REAR first conducts joint supervised optimization on reasoning-enhanced data, which serves as a warm-up to strengthen the Large Language Model (LLM)’s ability to perform EAE while incorporating auxiliary cues from RE. Subsequently, it applies reinforcement learning to explore diverse reasoning trajectories and derive near-optimal strategies for integrating relation-level signals into EAE. Experiments on the ACE-E, ACE-E⁺ and ERE benchmarks demonstrate that REAR consistently surpasses previous decoder-only LLM methods, achieving F1-score gains of at least 0.9%, 2.2% and 1.6%, respectively.

We introduce DA-Code, a code generation benchmark specifically designed to assess LLMs on agent-based data science tasks. This benchmark features three core elements: First, the tasks within DA-Code are inherently challenging, setting them apart from traditional code generation tasks and demanding advanced coding skills in grounding and planning. Second, examples in DA-Code are all based on real and diverse data, covering a wide range of complex data wrangling and analytics tasks. Third, to solve the tasks, the models must utilize complex data science programming languages, including Python and SQL, to perform intricate data processing and derive the answers. We set up the benchmark in a controllable and executable environment that aligns with real-world data analysis scenarios and is scalable. The annotators meticulously designed the evaluation suite to ensure the accuracy and robustness of the evaluation. We developed the DA-Agent baseline. Experiments show that although the baseline performs better than other existing frameworks, using the current best LLMs achieves only 30.5% accuracy, leaving ample room for improvement. We release our benchmark at [link](https://github.com/yiyihum/dabench)