Jialong Chen

YanMing Hu, TianChi Liao, JiaLong Chen et al.

Graph Neural networks (GNNs) have been applied in many scenarios due to the superior performance of graph learning. However, fairness is always ignored when designing GNNs. As a consequence, biased information in training data can easily affect vanilla GNNs, causing biased results toward particular demographic groups (divided by sensitive attributes, such as race and age). There have been efforts to address the fairness issue. However, existing fair techniques generally divide the demographic groups by raw sensitive attributes and assume that are fixed. The biased information correlated with raw sensitive attributes will run through the training process regardless of the implemented fair techniques. It is urgent to resolve this problem for training fair GNNs. To tackle this problem, we propose a brand new framework, FairMigration, which can dynamically migrate the demographic groups instead of keeping that fixed with raw sensitive attributes. FairMigration is composed of two training stages. In the first stage, the GNNs are initially optimized by personalized self-supervised learning, and the demographic groups are adjusted dynamically. In the second stage, the new demographic groups are frozen and supervised learning is carried out under the constraints of new demographic groups and adversarial training. Extensive experiments reveal that FairMigration balances model performance and fairness well.

Jialong Chen, Xander Xu, Hu Wei et al.

Large language model (LLM)-powered agents have demonstrated strong capabilities in automating software engineering tasks such as static bug fixing, as evidenced by benchmarks like SWE-bench. However, in the real world, the development of mature software is typically predicated on complex requirement changes and long-term feature iterations -- a process that static, one-shot repair paradigms fail to capture. To bridge this gap, we propose \textbf{SWE-CI}, the first repository-level benchmark built upon the Continuous Integration loop, aiming to shift the evaluation paradigm for code generation from static, short-term \textit{functional correctness} toward dynamic, long-term \textit{maintainability}. The benchmark comprises 100 tasks, each corresponding on average to an evolution history spanning 233 days and 71 consecutive commits in a real-world code repository. SWE-CI requires agents to systematically resolve these tasks through dozens of rounds of analysis and coding iterations. SWE-CI provides valuable insights into how well agents can sustain code quality throughout long-term evolution.

Weiqi Zhai, Zhihai Wang, Jinghang Wang et al.

Humanity's Last Exam (HLE) has become a widely used benchmark for evaluating frontier large language models on challenging, multi-domain questions. However, community-led analyses have raised concerns that HLE contains a non-trivial number of noisy items, which can bias evaluation results and distort cross-model comparisons. To address this challenge, we introduce HLE-Verified, a verified and revised version of HLE with a transparent verification protocol and fine-grained error taxonomy. Our construction follows a two-stage validation-and-repair workflow resulting in a certified benchmark. In Stage I, each item undergoes binary validation of the problem and final answer through domain-expert review and model-based cross-checks, yielding 641 verified items. In Stage II, flawed but fixable items are revised under strict constraints preserving the original evaluation intent, through dual independent expert repairs, model-assisted auditing, and final adjudication, resulting in 1,170 revised-and-certified items. The remaining 689 items are released as a documented uncertain set with explicit uncertainty sources and expertise tags for future refinement. We evaluate seven state-of-the-art language models on HLE and HLE-Verified, observing an average absolute accuracy gain of 7--10 percentage points on HLE-Verified. The improvement is particularly pronounced on items where the original problem statement and/or reference answer is erroneous, with gains of 30--40 percentage points. Our analyses further reveal a strong association between model confidence and the presence of errors in the problem statement or reference answer, supporting the effectiveness of our revisions. Overall, HLE-Verified improves HLE-style evaluations by reducing annotation noise and enabling more faithful measurement of model capabilities. Data is available at: https://github.com/SKYLENAGE-AI/HLE-Verified