Dissecting Physics Reasoning in Small Language Models: A Multi-Dimensional Analysis from an Educational Perspective

Small Language Models (SLMs) offer computational efficiency and accessibility, making them promising for educational applications. However, their capacity for complex reasoning, particularly in domains such as physics, remains underexplored. This study investigates the high school physics reasoning capabilities of state-of-the-art SLMs (under 4 billion parameters), including instruct versions of Llama 3.2, Phi 4 Mini, Gemma 3, and Qwen series. We developed a comprehensive physics dataset from the OpenStax High School Physics textbook, annotated according to Bloom's Taxonomy, with LaTeX and plaintext mathematical notations. A novel cultural contextualization approach was applied to a subset, creating culturally adapted problems for Asian, African, and South American/Australian contexts while preserving core physics principles. Using an LLM-as-a-judge framework with Google's Gemini 2.5 Flash, we evaluated answer and reasoning chain correctness, along with calculation accuracy. The results reveal significant differences between the SLMs. Qwen 3 1.7B achieved high `answer accuracy' (85%), but `fully correct reasoning' was substantially low (38%). The format of the mathematical notation had a negligible impact on performance. SLMs exhibited varied performance across the physics topics and showed a decline in reasoning quality with increasing cognitive and knowledge complexity. In particular, the consistency of reasoning was largely maintained in diverse cultural contexts, especially by better performing models. These findings indicate that, while SLMs can often find correct answers, their underlying reasoning is frequently flawed, suggesting an overreliance on pattern recognition. For SLMs to become reliable educational tools in physics, future development must prioritize enhancing genuine understanding and the generation of sound, verifiable reasoning chains over mere answer accuracy.

Small Language Models in the Real World: Insights from Industrial Text Classification

With the emergence of ChatGPT, Transformer models have significantly advanced text classification and related tasks. Decoder-only models such as Llama exhibit strong performance and flexibility, yet they suffer from inefficiency on inference due to token-by-token generation, and their effectiveness in text classification tasks heavily depends on prompt quality. Moreover, their substantial GPU resource requirements often limit widespread adoption. Thus, the question of whether smaller language models are capable of effectively handling text classification tasks emerges as a topic of significant interest. However, the selection of appropriate models and methodologies remains largely underexplored. In this paper, we conduct a comprehensive evaluation of prompt engineering and supervised fine-tuning methods for transformer-based text classification. Specifically, we focus on practical industrial scenarios, including email classification, legal document categorization, and the classification of extremely long academic texts. We examine the strengths and limitations of smaller models, with particular attention to both their performance and their efficiency in Video Random-Access Memory (VRAM) utilization, thereby providing valuable insights for the local deployment and application of compact models in industrial settings.

Delving into Multilingual Ethical Bias: The MSQAD with Statistical Hypothesis Tests for Large Language Models

Despite the recent strides in large language models, studies have underscored the existence of social biases within these systems. In this paper, we delve into the validation and comparison of the ethical biases of LLMs concerning globally discussed and potentially sensitive topics, hypothesizing that these biases may arise from language-specific distinctions. Introducing the Multilingual Sensitive Questions & Answers Dataset (MSQAD), we collected news articles from Human Rights Watch covering 17 topics, and generated socially sensitive questions along with corresponding responses in multiple languages. We scrutinized the biases of these responses across languages and topics, employing two statistical hypothesis tests. The results showed that the null hypotheses were rejected in most cases, indicating biases arising from cross-language differences. It demonstrates that ethical biases in responses are widespread across various languages, and notably, these biases were prevalent even among different LLMs. By making the proposed MSQAD openly available, we aim to facilitate future research endeavors focused on examining cross-language biases in LLMs and their variant models.

Internal and External Impacts of Natural Language Processing Papers

We investigate the impacts of NLP research published in top-tier conferences (i.e., ACL, EMNLP, and NAACL) from 1979 to 2024. By analyzing citations from research articles and external sources such as patents, media, and policy documents, we examine how different NLP topics are consumed both within the academic community and by the broader public. Our findings reveal that language modeling has the widest internal and external influence, while linguistic foundations have lower impacts. We also observe that internal and external impacts generally align, but topics like ethics, bias, and fairness show significant attention in policy documents with much fewer academic citations. Additionally, external domains exhibit distinct preferences, with patents focusing on practical NLP applications and media and policy documents engaging more with the societal implications of NLP models.

DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural Language and Reinforcement Learning

Theorem proving serves as a major testbed for evaluating complex reasoning abilities in large language models (LLMs). However, traditional automated theorem proving (ATP) approaches rely heavily on formal proof systems that poorly align with LLMs' strength derived from informal, natural language knowledge acquired during pre-training. In this work, we propose DeepTheorem, a comprehensive informal theorem-proving framework exploiting natural language to enhance LLM mathematical reasoning. DeepTheorem includes a large-scale benchmark dataset consisting of 121K high-quality IMO-level informal theorems and proofs spanning diverse mathematical domains, rigorously annotated for correctness, difficulty, and topic categories, accompanied by systematically constructed verifiable theorem variants. We devise a novel reinforcement learning strategy (RL-Zero) explicitly tailored to informal theorem proving, leveraging the verified theorem variants to incentivize robust mathematical inference. Additionally, we propose comprehensive outcome and process evaluation metrics examining proof correctness and the quality of reasoning steps. Extensive experimental analyses demonstrate DeepTheorem significantly improves LLM theorem-proving performance compared to existing datasets and supervised fine-tuning protocols, achieving state-of-the-art accuracy and reasoning quality. Our findings highlight DeepTheorem's potential to fundamentally advance automated informal theorem proving and mathematical exploration.

Dynamic Malware Classification of Windows PE Files using CNNs and Greyscale Images Derived from Runtime API Call Argument Conversion

Malware detection and classification remains a topic of concern for cybersecurity, since it is becoming common for attackers to use advanced obfuscation on their malware to stay undetected. Conventional static analysis is not effective against polymorphic and metamorphic malware as these change their appearance without modifying their behavior, thus defying the analysis by code structure alone. This makes it important to use dynamic detection that monitors malware behavior at runtime. In this paper, we present a dynamic malware categorization framework that extracts API argument calls at the runtime execution of Windows Portable Executable (PE) files. Extracting and encoding the dynamic features of API names, argument return values, and other relative features, we convert raw behavioral data to temporal patterns. To enhance feature portrayal, the generated patterns are subsequently converted into grayscale pictures using a magma colormap. These improved photos are used to teach a Convolutional Neural Network (CNN) model discriminative features, which allows for reliable and accurate malware classification. Results from experiments indicate that our method, with an average accuracy of 98.36% is effective in classifying different classes of malware and benign by integrating dynamic analysis and deep learning. It not only achieves high classification accuracy but also demonstrates significant resilience against typical evasion strategies.

Robust LLM Fingerprinting via Domain-Specific Watermarks

As open-source language models (OSMs) grow more capable and are widely shared and finetuned, ensuring model provenance, i.e., identifying the origin of a given model instance, has become an increasingly important issue. At the same time, existing backdoor-based model fingerprinting techniques often fall short of achieving key requirements of real-world model ownership detection. In this work, we build on the observation that while current open-source model watermarks fail to achieve reliable content traceability, they can be effectively adapted to address the challenge of model provenance. To this end, we introduce the concept of domain-specific watermarking for model fingerprinting. Rather than watermarking all generated content, we train the model to embed watermarks only within specified subdomains (e.g., particular languages or topics). This targeted approach ensures detection reliability, while improving watermark durability and quality under a range of real-world deployment settings. Our evaluations show that domain-specific watermarking enables model fingerprinting with strong statistical guarantees, controllable false positive rates, high detection power, and preserved generation quality. Moreover, we find that our fingerprints are inherently stealthy and naturally robust to real-world variability across deployment scenarios.

Optimizing Image Capture for Computer Vision-Powered Taxonomic Identification and Trait Recognition of Biodiversity Specimens

Biological collections house millions of specimens documenting Earth's biodiversity, with digital images increasingly available through open-access platforms. Most imaging protocols were developed for human visual interpretation without considering computational analysis requirements. This paper aims to bridge the gap between current imaging practices and the potential for automated analysis by presenting key considerations for creating biological specimen images optimized for computer vision applications. We provide conceptual computer vision topics for context, addressing fundamental concerns including model generalization, data leakage, and comprehensive metadata documentation, and outline practical guidance on specimen imagine, and data storage. These recommendations were synthesized through interdisciplinary collaboration between taxonomists, collection managers, ecologists, and computer scientists. Through this synthesis, we have identified ten interconnected considerations that form a framework for successfully integrating biological specimen images into computer vision pipelines. The key elements include: (1) comprehensive metadata documentation, (2) standardized specimen positioning, (3) consistent size and color calibration, (4) protocols for handling multiple specimens in one image, (5) uniform background selection, (6) controlled lighting, (7) appropriate resolution and magnification, (8) optimal file formats, (9) robust data archiving strategies, and (10) accessible data sharing practices. By implementing these recommendations, collection managers, taxonomists, and biodiversity informaticians can generate images that support automated trait extraction, species identification, and novel ecological and evolutionary analyses at unprecedented scales. Successful implementation lies in thorough documentation of methodological choices.

Who You Are Matters: Bridging Topics and Social Roles via LLM-Enhanced Logical Recommendation

Recommender systems filter contents/items valuable to users by inferring preferences from user features and historical behaviors. Mainstream approaches follow the learning-to-rank paradigm, which focus on discovering and modeling item topics (e.g., categories), and capturing user preferences on these topics based on historical interactions. However, this paradigm often neglects the modeling of user characteristics and their social roles, which are logical confounders influencing the correlated interest and user preference transition. To bridge this gap, we introduce the user role identification task and the behavioral logic modeling task that aim to explicitly model user roles and learn the logical relations between item topics and user social roles. We show that it is possible to explicitly solve these tasks through an efficient integration framework of Large Language Model (LLM) and recommendation systems, for which we propose TagCF. On the one hand, the exploitation of the LLM's world knowledge and logic inference ability produces a virtual logic graph that reveals dynamic and expressive knowledge of users, augmenting the recommendation performance. On the other hand, the user role aligns the user behavioral logic with the observed user feedback, refining our understanding of user behaviors. Additionally, we also show that the extracted user-item logic graph is empirically a general knowledge that can benefit a wide range of recommendation tasks, and conduct experiments on industrial and several public datasets as verification.

R1dacted: Investigating Local Censorship in DeepSeek's R1 Language Model

DeepSeek recently released R1, a high-performing large language model (LLM) optimized for reasoning tasks. Despite its efficient training pipeline, R1 achieves competitive performance, even surpassing leading reasoning models like OpenAI's o1 on several benchmarks. However, emerging reports suggest that R1 refuses to answer certain prompts related to politically sensitive topics in China. While existing LLMs often implement safeguards to avoid generating harmful or offensive outputs, R1 represents a notable shift - exhibiting censorship-like behavior on politically charged queries. In this paper, we investigate this phenomenon by first introducing a large-scale set of heavily curated prompts that get censored by R1, covering a range of politically sensitive topics, but are not censored by other models. We then conduct a comprehensive analysis of R1's censorship patterns, examining their consistency, triggers, and variations across topics, prompt phrasing, and context. Beyond English-language queries, we explore censorship behavior in other languages. We also investigate the transferability of censorship to models distilled from the R1 language model. Finally, we propose techniques for bypassing or removing this censorship. Our findings reveal possible additional censorship integration likely shaped by design choices during training or alignment, raising concerns about transparency, bias, and governance in language model deployment.