Yan Tao

Yan Tao, Olga Viberg, Ryan S. Baker et al.

Culture fundamentally shapes people's reasoning, behavior, and communication. As people increasingly use generative artificial intelligence (AI) to expedite and automate personal and professional tasks, cultural values embedded in AI models may bias people's authentic expression and contribute to the dominance of certain cultures. We conduct a disaggregated evaluation of cultural bias for five widely used large language models (OpenAI's GPT-4o/4-turbo/4/3.5-turbo/3) by comparing the models' responses to nationally representative survey data. All models exhibit cultural values resembling English-speaking and Protestant European countries. We test cultural prompting as a control strategy to increase cultural alignment for each country/territory. For recent models (GPT-4, 4-turbo, 4o), this improves the cultural alignment of the models' output for 71-81% of countries and territories. We suggest using cultural prompting and ongoing evaluation to reduce cultural bias in the output of generative AI.

Tingcheng Bian, Jinchang Luo, Mingquan Cheng et al.

Large language models achieve breakthroughs in complex reasoning via long chain-of-thought sequences. However, this often leads to severe reasoning inflation, causing substantial computational redundancy. To maximize Intelligence per Token, we introduce a theoretical metric, MSL-Minimal Sufficient Length. MSL rigorously characterizes the shortest reasoning length that preserves answer correctness. We provide a recursive definition based on independently sampled sequences and prove the existence of its limit, establishing the first measurable lower bound for reasoning-chain compression. Building on an analysis of mainstream CoT compression strategies, we identify key structural factors enabling a model to approach MSL. Based on these insights, we propose TRiMS which employs the GRPO algorithm in conjunction with MSL-based estimation during training, while mitigating instabilities during the training process through dynamic batch aggregation and advantage computation using batch-level standard deviation. TRiMS achieves over 80% CoT token reduction with a minor accuracy boost across all benchmarks.

Yan Tao, Olga Viberg, Deepak Varuvel Dennison et al.

Teachers' trust in artificial intelligence (AI) in education depends on how they balance its perceived benefits and risks. Yet global discussions about scaling AI in education rely on fragmented evidence, as most studies of teachers' perceptions focus on single countries or small samples. This lack of representative cross-national evidence limits both theory building and policy development. At the same time, large language models (LLMs) are increasingly used in research, policy, and teachers' professional workflows, despite limited validation in education. To address these gaps, we conduct a large-scale audit of LLM alignment with teachers' perceptions of AI by combining representative international survey data with systematic model evaluation. Using OECD TALIS data from 55 countries and territories, we measure cross-national variation in teachers' perceived benefits and risks of AI. We then benchmark responses from eight state-of-the-art LLMs across four providers under both general and country-specific prompting, comparing higher- and lower-reasoning models. Results reveal substantial cross-national variation in teacher perceptions that is not reliably reflected in LLM outputs. Models compress country differences, overestimate both benefits and risks, and show limited gains from identity prompting or enhanced reasoning. This misalignment matters because LLM-generated guidance and professional discourse increasingly shape how teachers learn about and discuss AI, potentially influencing trust and future adoption decisions. Our findings caution against treating LLM outputs as substitutes for direct engagement with teachers when informing global AI-in-education initiatives. At the same time, some models (e.g., Gemini 3 Fast) partially capture cross-national ranking patterns, suggesting a complementary role in hypothesis generation and exploratory comparative analysis.

Jinchang Luo, Mingquan Cheng, Fan Wan et al.

Reinforcement learning has recently shown promise in improving retrieval-augmented generation (RAG). Despite these advances, its effectiveness in multi-hop question answering (QA) remains limited by two fundamental limitations: (i) global planning absence to structure multi-step reasoning, and (ii) unfaithful execution, which hinders effective query formulation and consistent use of retrieved evidence. We propose GlobalRAG, a reinforcement learning framework designed to enhance global reasoning in multi-hop QA. GlobalRAG decomposes questions into subgoals, coordinates retrieval with reasoning, and refines evidence iteratively. To guide this process, we introduce Planning Quality Reward and SubGoal Completion Reward, which encourage coherent planning and reliable subgoal execution. In addition, a progressive weight annealing strategy balances process-oriented and outcome-based objectives. Extensive experiments on both in-domain and out-of-domain benchmarks demonstrate that GlobalRAG significantly outperforms strong baselines while using only 8k training data (42% of the training data used by strong baselines), achieving average improvements of 14.2% in both EM and F1.