Practical guidance on training Large Language Models (LLMs) to leverage Code Interpreter across diverse tasks remains lacking. Our final model, R1-CI-14B, improves average accuracy on the 37 test tasks from 44.1% to 72.4%, Notably, R1-CI-14B also exhibits emergent self-checking behavior through code generation. While reinforcement learning (RL)-based fine-tuning has significantly improved LLMs' reasoning and planning Wang In contrast, symbolic code generation handles these rigorously and benefits from external tools (e.g., A key challenge is guiding LLMs to decide when to rely on textual reasoning versus programmatic solutions, given that most input questions lack explicit cues about which approach is best and the possible text/code solution space is large. OpenAI's GPT models address this by incorporating a Code Interpreter, allowing iterative code generation Interpreter implementations struggle to effectively steer between text and code, underutilizing symbolic capabilities. Recent work such as ToRL (Li et al., 2025b) and ReTool (Feng et al., 2025) investigates training reasoning models to integrate with Code Interpreters. To tackle these challenges, we present R1-Code-Interpreter, a framework for integrating Code Interpreter into open-source LLMs. We curate 144 reasoning and planning tasks and synthesize 6.5k multi-turn text/code trajectories for This difficulty arises from task heterogeneity and the scarcity of effective samples.

Thirty years ago, a peace-loving Austrian theologian spoke to Peter Thiel about the apocalyptic theories of Nazi jurist Carl Schmitt. They've been a road map for the billionaire ever since. For a full two years now, the billionaire has been on the circuit, spreading his biblically inflected ideas about doomsday through a set of variably and sometimes visibly perplexed interviewers. He has chatted onstage with the economist podcaster Tyler Cowen about the (the scriptural term for "that which withholds" the end times); traded some very awkward on-camera silences with the New York Times columnist Ross Douthat; and is, at this very moment, in the midst of delivering a four-part, off-the-record lecture series about the Antichrist in San Francisco. Depending on who you are, you may find it hilarious, fascinating, insufferable, or horrifying that one of the world's most powerful men is obsessing over a figure from sermons and horror movies. But the ideas and influences behind these talks are key to understanding how Thiel sees his own massive role in the world--in politics, technology, and the fate of the species. And to really grasp Thiel's katechon-and-Antichrist schtick, you need to go back to the first major lecture of his doomsday road show--which took place on an unusually hot day in Paris in 2023. No video cameras recorded the event, and no reporters wrote about it, but I've been able to reconstruct it by talking to people who were there. The venue was a yearly conference of scholars devoted to Thiel's chief intellectual influence, the late French-American theorist René Girard. On the evening of the unpublicized lecture, dozens of Girardian philosophers and theologians from around the world filed into a modest lecture hall at the Catholic University of Paris. And from the dais, Thiel delivered a nearly hourlong account of his thoughts on Armageddon--and all the things he believed were "not enough" to prevent it. By Thiel's telling, the modern world is scared, way too scared, of its own technology. Our "listless" and "zombie" age, he said, is marked by a growing hostility to innovation, plummeting fertility rates, too much yoga, and a culture mired in the "endless Groundhog Day of the worldwide web." But in its neurotic desperation to avoid technological Armageddon--the real threats of nuclear war, environmental catastrophe, runaway AI--modern civilization has become susceptible to something even more dangerous: the Antichrist. According to some Christian traditions, the Antichrist is a figure that will unify humanity under one rule before delivering us to the apocalypse. For Thiel, its evil is pretty much synonymous with any attempt to unite the world. "How might such an Antichrist rise to power?" Thiel asked.

Recently, online Reinforcement Learning with Verifiable Rewards (RLVR) has become a key paradigm for enhancing the reasoning capabilities of Large Language Models (LLMs). However, existing methods typically treat all training samples uniformly, overlooking the vast differences in problem difficulty relative to the model's current capabilities. This uniform training strategy leads to inefficient exploration of problems the model has already mastered, while concurrently lacking effective guidance on problems that are challenging its abilities the most, limiting both learning efficiency and upper-bound performance. To address this, we propose CLPO (Curriculum-guided Learning for Policy Optimization), a novel algorithm that creates a dynamic pedagogical feedback loop within the policy optimization process. The core of CLPO leverages the model's own rollout performance to conduct real-time difficulty assessment, thereby constructing an Online Curriculum. This curriculum then guides an Adaptive Problem Restructuring mechanism, where the model acts as its own teacher: it diversifies medium-difficulty problems to promote generalization and simplifies challenging problems to make them more attainable. Our approach transforms the static training procedure into a dynamic process that co-evolves with the model's capabilities. Experiments show that CLPO achieves state-of-the-art performance across eight challenging mathematical and general reasoning benchmarks, with an average pass@1 improvement of 6.96% over other methods, demonstrating its potential for more efficiently training more capable reasoning models.

GUI agents powered by vision-language models (VLMs) show promise in automating complex digital tasks. However, their effectiveness in real-world applications is often limited by scarce training data and the inherent complexity of these tasks, which frequently require long-tailed knowledge covering rare, unseen scenarios. We propose RAG-GUI , a lightweight VLM that leverages web tutorials at inference time. RAG-GUI is first warm-started via supervised finetuning (SFT) and further refined through self-guided rejection sampling finetuning (RSF). Designed to be model-agnostic, RAG-GUI functions as a generic plug-in that enhances any VLM-based agent. Evaluated across three distinct tasks, it consistently outperforms baseline agents and surpasses other inference baselines by 2.6% to 13.3% across two model sizes, demonstrating strong generalization and practical plug-and-play capabilities in real-world scenarios.

As digital platforms redefine educational paradigms, ensuring interactivity remains vital for effective learning. This paper explores using Multimodal Large Language Models (MLLMs) to automatically respond to student questions from online lectures - a novel question answering task of real world significance. We introduce the EduVidQA Dataset with 5252 question-answer pairs (both synthetic and real-world) from 296 computer science videos covering diverse topics and difficulty levels. To understand the needs of the dataset and task evaluation, we empirically study the qualitative preferences of students, which we provide as an important contribution to this line of work. Our benchmarking experiments consist of 6 state-of-the-art MLLMs, through which we study the effectiveness of our synthetic data for finetuning, as well as showing the challenging nature of the task. We evaluate the models using both text-based and qualitative metrics, thus showing a nuanced perspective of the models' performance, which is paramount to future work. This work not only sets a benchmark for this important problem, but also opens exciting avenues for future research in the field of Natural Language Processing for Education.

Learning to learn is becoming a science, driven by the convergence of knowledge tracing, signal processing, and generative AI to model student learning states and optimize education. We propose CoTutor, an AI-driven model that enhances Bayesian Knowledge Tracing with signal processing techniques to improve student progress modeling and deliver adaptive feedback and strategies. Deployed as an AI copilot, CoTutor combines generative AI with adaptive learning technology. In university trials, it has demonstrated measurable improvements in learning outcomes while outperforming conventional educational tools. Our results highlight its potential for AI-driven personalization, scalability, and future opportunities for advancing privacy and ethical considerations in educational technology. Inspired by Richard Hamming's vision of computer-aided 'learning to learn,' CoTutor applies convex optimization and signal processing to automate and scale up learning analytics, while reserving pedagogical judgment for humans, ensuring AI facilitates the process of knowledge tracing while enabling learners to uncover new insights.

Artificial intelligence (AI) is reshaping higher education, yet current debates often feel tangled, mixing concerns about pedagogy, operations, curriculum, and the future of work without a shared framework. This paper offers a first attempt at a taxonomy to organize the diverse narratives of AI education and to inform discipline-based curricular discussions. We place these narratives within the enduring responsibility of higher education: the mission of knowledge. This mission includes not only the preservation and advancement of disciplinary expertise, but also the cultivation of skills and wisdom, i.e., forms of meta-knowledge that encompass judgment, ethics, and social responsibility. For the purpose of this paper's discussion, AI is defined as adaptive, data-driven systems that automate analysis, modeling, and decision-making, highlighting its dual role as enabler and disruptor across disciplines. We argue that the most consequential challenges lie at the level of curriculum and disciplinary purpose, where AI accelerates inquiry but also unsettles expertise and identity. We show how disciplines evolve through the interplay of research, curriculum, pedagogy, and faculty expertise, and why curricular reform is the central lever for meaningful change. Pedagogical innovation offers a strategic and accessible entry point, providing actionable steps that help faculty and students build the expertise needed to engage in deeper curricular rethinking and disciplinary renewal. Within this framing, we suggest that meaningful reform can move forward through structured faculty journeys: from AI literacy to pedagogy, curriculum design, and research integration. The key is to align these journeys with the mission of knowledge, turning the disruptive pressures of AI into opportunities for disciplines to sustain expertise, advance inquiry, and serve society.

In-context Learning (ICL) utilizes structured demonstration-query inputs to induce few-shot learning on Language Models (LMs), which are not originally pre-trained on ICL-style data. To bridge the gap between ICL and pre-training, some approaches fine-tune LMs on large ICL-style datasets by an end-to-end paradigm with massive computational costs. To reduce such costs, in this paper, we propose Attention Behavior Fine-Tuning (ABFT), utilizing the previous findings on the inner mechanism of ICL, building training objectives on the attention scores instead of the final outputs, to force the attention scores to focus on the correct label tokens presented in the context and mitigate attention scores from the wrong label tokens. Our experiments on 9 modern LMs and 8 datasets empirically find that ABFT outperforms in performance, robustness, unbiasedness, and efficiency, with only around 0.01% data cost compared to the previous methods. Moreover, our subsequent analysis finds that the end-to-end training objective contains the ABFT objective, suggesting the implicit bias of ICL-style data to the emergence of induction heads. Our work demonstrates the possibility of controlling specific module sequences within LMs to improve their behavior, opening up the future application of mechanistic interpretability.

Test-time Scaling (TTS) has been demonstrated to significantly enhance the reasoning capabilities of Large Language Models (LLMs) during the inference phase without altering model parameters. However, existing TTS methods are largely independent, implying that LLMs have not yet evolved to progressively learn how to scale more effectively. With the objective of evolving LLMs to learn "how to scale test-time computation," we propose LatentEvolve, a self-evolving latent TTS framework inspired by the complementary learning system (CLS) theory. Analogous to the human brain's dual system of a fast-recall hippocampus and a slow-consolidating neocortex, LatentEvolve comprises two evolutionary components: daytime scaling, which rapidly retrieves historical latent representations to better guide current LLM reasoning; and nighttime scaling, which integrates past latent optimizations in a manner akin to the human brain's consolidation of experiences during sleep. The alternation of daytime and nighttime processes facilitates a fast and slow evolution of LLM TTS, mirroring human cognitive dynamics in a fully unsupervised manner. Extensive experiments across eight benchmarks and five model backbones demonstrate that our LatentEvolve surpasses state-of-the-art TTS methods such as LatentSeek and TTRL by up to 13.33% and exhibits exceptional cross-domain and cross-backbone generalization. Much of this success in recent years has been driven by training-time scaling, wherein increasing the volume of training data and parameters consistently yields performance improvements (Kaplan et al., 2020; Aghajanyan et al., 2023). However, the pace of this scaling, particularly in terms of pre-training scale, has begun to slow, constrained by its resource-intensive nature and the depletion of high-quality training data (Villalobos et al., 2022; Zhou et al., 2025).

Humans quickly learn new concepts from a small number of examples. Replicating this capacity with Artificial Intelligence (AI) systems has proven to be challenging. When it comes to learning subjective tasks-where there is an evident scarcity of data-this capacity needs to be recreated. In this work, we propose an intuitive human-agent teaching architecture in which the human can teach an agent how to perform a task by providing demonstrations, i.e., examples. To have an intuitive interaction, we argue that the agent should be able to learn incrementally from a few single examples. To allow for this, our objective is to broaden the agent's task understanding using domain knowledge. Then, using a learning method to enable the agent to learn efficiently from a limited number of examples. Finally, to optimize how human can select the most representative and less redundant examples to provide the agent with. We apply our proposed method to the subjective task of ingredient substitution, where the agent needs to learn how to substitute ingredients in recipes based on human examples. We replicate human input using the Recipe1MSubs dataset. In our experiments, the agent achieves half its task performance after only 100 examples are provided, compared to the complete training set of 50k examples. We show that by providing examples in strategic order along with a learning method that leverages external symbolic knowledge, the agent can generalize more efficiently.