Reinforcement learning (RL) has recently become a strong recipe for training reasoning LLMs that produce long chains of thought (LongCoT). Yet the standard RL "thinking environment", where the state is the prompt plus all prior reasoning tokens, makes the state unbounded and forces attention-based policies to pay quadratic compute as thoughts lengthen. We revisit the environment itself. We propose Markovian Thinking, a paradigm in which the policy advances reasoning while conditioning on a constant-size state, decoupling thinking length from context size. As an immediate consequence this yields linear compute with constant memory. We instantiate this idea with Delethink, an RL environment that structures reasoning into fixed-size chunks. Within each chunk, the model thinks as usual; at the boundary, the environment resets the context and reinitializes the prompt with a short carryover. Through RL, the policy learns to write a textual state near the end of each chunk sufficient for seamless continuation of reasoning after reset. Trained in this environment, an R1-Distill 1.5B model reasons in 8K-token chunks yet thinks up to 24K tokens, matching or surpassing LongCoT-RL trained with a 24K budget. With test-time scaling, Delethink continues to improve where LongCoT plateaus. The effect of linear compute is substantial: we empirically estimate at 96K average thinking length LongCoT-RL costs 27 H100-months vs. 7 for Delethink. Analysis at RL initialization shows off-the-shelf reasoning models (1.5B-120B) often sample Markovian traces zero-shot across diverse benchmarks, providing positive samples that make RL effective at scale. Our results show that redesigning the thinking environment is a powerful lever: it enables very long reasoning without quadratic overhead and opens a path toward efficient, scalable reasoning LLMs.

We investigate whether contemporary multimodal LLMs can assist with grading open-ended calculus at scale without eroding validity. In a large first-year exam, students' handwritten work was graded by GPT-5 against the same rubric used by teaching assistants (TAs), with fractional credit permitted; TA rubric decisions served as ground truth. We calibrated a human-in-the-loop filter that combines a partial-credit threshold with an Item Response Theory (2PL) risk measure based on the deviation between the AI score and the model-expected score for each student-item. Unfiltered AI-TA agreement was moderate, adequate for low-stakes feedback but not for high-stakes use. Confidence filtering made the workload-quality trade-off explicit: under stricter settings, AI delivered human-level accuracy, but also left roughly 70% of the items to be graded by humans. Psychometric patterns were constrained by low stakes on the open-ended portion, a small set of rubric checkpoints, and occasional misalignment between designated answer regions and where work appeared. Practical adjustments such as slightly higher weight and protected time, a few rubric-visible substeps, stronger spatial anchoring should raise ceiling performance. Overall, calibrated confidence and conservative routing enable AI to reliably handle a sizable subset of routine cases while reserving expert judgment for ambiguous or pedagogically rich responses.

The superposition hypothesis states that single neurons may participate in representing multiple features in order for the neural network to represent more features than it has neurons. In neuroscience and AI, representational alignment metrics measure the extent to which different deep neural networks (DNNs) or brains represent similar information. In this work, we explore a critical question: does superposition interact with alignment metrics in any undesirable way? We hypothesize that models which represent the same features in different superposition arrangements, i.e., their neurons have different linear combinations of the features, will interfere with predictive mapping metrics (semi-matching, soft-matching, linear regression), producing lower alignment than expected. We develop a theory for how permutation metrics are dependent on superposition arrangements. This is tested by training sparse autoencoders (SAEs) to disentangle superposition in toy models, where alignment scores are shown to typically increase when a model's base neurons are replaced with its sparse overcomplete latent codes. We find similar increases for DNN-DNN and DNN-brain linear regression alignment in the visual domain. Our results suggest that superposition disentanglement is necessary for mapping metrics to uncover the true representational alignment between neural networks.

Large Language Models (LLMs) have demonstrated remarkable success across a wide range of industries, primarily due to their impressive generative abilities. Yet, their potential in applications requiring cognitive abilities, such as psychological counseling, remains largely untapped. This paper investigates the key question: \textit{Can LLMs be effectively applied to psychological counseling?} To determine whether an LLM can effectively take on the role of a psychological counselor, the first step is to assess whether it meets the qualifications required for such a role, namely the ability to pass the U.S. National Counselor Certification Exam (NCE). This is because, just as a human counselor must pass a certification exam to practice, an LLM must demonstrate sufficient psychological knowledge to meet the standards required for such a role. To address this, we introduce PsychCounsel-Bench, a benchmark grounded in U.S.national counselor examinations, a licensure test for professional counselors that requires about 70\% accuracy to pass. PsychCounsel-Bench comprises approximately 2,252 carefully curated single-choice questions, crafted to require deep understanding and broad enough to cover various sub-disciplines of psychology. This benchmark provides a comprehensive assessment of an LLM's ability to function as a counselor. Our evaluation shows that advanced models such as GPT-4o, Llama3.3-70B, and Gemma3-27B achieve well above the passing threshold, while smaller open-source models (e.g., Qwen2.5-7B, Mistral-7B) remain far below it. These results suggest that only frontier LLMs are currently capable of meeting counseling exam standards, highlighting both the promise and the challenges of developing psychology-oriented LLMs. We release the proposed dataset for public use: https://github.com/cloversjtu/PsychCounsel-Bench

Large Language Models (LLMs) are trained on massive, unstructured corpora, making it unclear which social patterns and biases they absorb and later reproduce. Existing evaluations typically examine outputs or activations, but rarely connect them back to the pre-training data. We introduce a pipeline that couples LLMs with sparse autoencoders (SAEs) to trace how different themes are encoded during training. As a controlled case study, we trained a GPT-style model on 37 nineteenth-century novels by ten female authors, a corpus centered on themes such as gender, marriage, class, and morality. By applying SAEs across layers and probing with eleven social and moral categories, we mapped sparse features to human-interpretable concepts. The analysis revealed stable thematic backbones (most prominently around gender and kinship) and showed how associations expand and entangle with depth. More broadly, we argue that the LLM+SAEs pipeline offers a scalable framework for auditing how cultural assumptions from the data are embedded in model representations.

Large Language Models (LLMs) and their multimodal extensions are becoming increasingly popular. One common approach to enable multimodality is to cascade domain-specific encoders with an LLM, making the resulting model inherit vulnerabilities from all of its components. In this work, we present the first systematic study of audio backdoor attacks against speech language models. We demonstrate its effectiveness across four speech encoders and three datasets, covering four tasks: automatic speech recognition (ASR), speech emotion recognition, and gender and age prediction. The attack consistently achieves high success rates, ranging from 90.76% to 99.41%. To better understand how backdoors propagate, we conduct a component-wise analysis to identify the most vulnerable stages of the pipeline. Finally, we propose a fine-tuning-based defense that mitigates the threat of poisoned pretrained encoders. Large language models (LLMs) are increasingly extended to multimodal settings, processing combinations of text, images, video, and audio (DeepMind, 2023; Biadsy et al., 2023; Radford et al., 2021; Rajaa & Tushar, 2024). While powerful, these systems inherit vulnerabilities from each of their components. Among them are backdoor attacks, in which a model behaves normally on clean inputs but produces targeted outputs when a hidden trigger is present (Gu et al., 2017). Prior backdoor studies have largely focused on single-modality large language models (Xu et al., 2023; Y ao et al., 2024) or speech processing models (Zhai et al., 2021; Koffas et al., 2022), leaving open questions about how such attacks propagate in a cascaded speech language model.

Prompt-driven scene synthesis allows users to generate complete 3D environments from textual descriptions. Current text-to-scene methods often struggle with complex geometries and object transformations, and tend to show weak adherence to complex instructions. We address these limitations by introducing Reason-3D, a text-to-scene model powered by large reasoning models (LRMs). Reason-3D integrates object retrieval using captions covering physical, functional, and contextual attributes. Reason-3D then places the selected objects based on implicit and explicit layout constraints, and refines their positions with collision-aware spatial reasoning. Evaluated on instructions ranging from simple to complex indoor configurations, Reason-3D significantly outperforms previous methods in human-rated visual fidelity, adherence to constraints, and asset retrieval quality. Beyond its contribution to the field of text-to-scene generation, our work showcases the advanced spatial reasoning abilities of modern LRMs. Additionally, we release the codebase to further the research in object retrieval and placement with LRMs.

This paper introduces DOoM, a new open-source benchmark designed to assess the capabilities of language models in solving mathematics and physics problems in Russian. The benchmark includes problems of varying difficulty, ranging from school-level tasks to university Olympiad and entrance exam questions. In this paper we discuss the motivation behind its creation, describe dataset's structure and evaluation methodology, and present initial results from testing various models. Analysis of the results shows a correlation between model performance and the number of tokens used, and highlights differences in performance between mathematics and physics tasks.

Recurrent-depth transformers scale test-time compute by iterating latent computations before emitting tokens. We study the geometry of these iterates and argue for a simple, two-scale operational picture: (i) within a looped block, updates act as small-scale refinements; (ii) across consecutive blocks, states undergo a larger-scale drift. Across training, our measurements show that loop steps become smaller and increasingly orthogonal to one another, indicating better local modeling of fine structure rather than merely pushing in a single direction. These dynamics motivate an early-exit mechanism based on the model's second-order difference in step-size, which we show is superior in terms of performance, stability and time-efficiency, when compared to the KL-divergence exit strategy of Geiping et al. and its naive first-order counterpart.

Why should a clinician trust an Artificial Intelligence (AI) prediction? Despite the increasing accuracy of machine learning methods in medicine, the lack of transparency continues to hinder their adoption in clinical practice. In this work, we explore Kolmogorov-Arnold Networks (KANs) for clinical classification tasks on tabular data. In contrast to traditional neural networks, KANs are function-based architectures that offer intrinsic interpretability through transparent, symbolic representations. We introduce \emph{Logistic-KAN}, a flexible generalization of logistic regression, and \emph{Kolmogorov-Arnold Additive Model (KAAM)}, a simplified additive variant that delivers transparent, symbolic formulas. Unlike ``black-box'' models that require post-hoc explainability tools, our models support built-in patient-level insights, intuitive visualizations, and nearest-patient retrieval. Across multiple health datasets, our models match or outperform standard baselines, while remaining fully interpretable. These results position KANs as a promising step toward trustworthy AI that clinicians can understand, audit, and act upon. We release the code for reproducibility in \codeurl.