Large decoder-based language models have become the dominant architecture for reward modeling in reinforcement learning from human feedback (RLHF). However, as reward models are increasingly deployed in test-time strategies, their inference costs become a growing concern. We present TinyRM, a family of small, bidirectional masked language models (MLMs) with as few as 400 million parameters, that rival the capabilities of models over 175 times larger on reasoning and safety preference modeling tasks. TinyRM combines FLAN-style prompting, Directional Low-Rank Adaptation (DoRA), and layer freezing to achieve strong performance on RewardBench, despite using significantly fewer resources. Our experiments suggest that small models benefit from domain-specific tuning strategies, particularly in reasoning, where lightweight finetuning methods are especially effective. While challenges remain in building generalist models and conversational preference modeling, our preliminary results highlight the promise of lightweight bidirectional architectures as efficient, scalable alternatives for preference modeling.

Recent advancements have demonstrated the significant potential of large language models (LLMs) in analog circuit design. Nevertheless, testbench construction for analog circuits remains manual, creating a critical bottleneck in achieving fully automated design processes. Particularly when replicating circuit designs from academic papers, manual Testbench construction demands time-intensive implementation and frequent adjustments, which fails to address the dynamic diversity and flexibility requirements for automation. AnalogTester tackles automated analog design challenges through an LLM-powered pipeline: a) domain-knowledge integration, b) paper information extraction, c) simulation scheme synthesis, and d) testbench code generation with Tsinghua Electronic Design (TED). AnalogTester has demonstrated automated Testbench generation capabilities for three fundamental analog circuit types: operational amplifiers (op-amps), bandgap references (BGRs), and low-dropout regulators (LDOs), while maintaining a scalable framework for adaptation to broader circuit topologies. Furthermore, AnalogTester can generate circuit knowledge data and TED code corpus, establishing fundamental training datasets for LLM specialization in analog circuit design automation.

Retrieval-Augmented Generation (RAG) systems commonly use chunking strategies for retrieval, which enhance large language models (LLMs) by enabling them to access external knowledge, ensuring that the retrieved information is up-to-date and domain-specific. However, traditional methods often fail to create chunks that capture sufficient semantic meaning, as they do not account for the underlying textual structure. This paper proposes a novel framework that enhances RAG by integrating hierarchical text segmentation and clustering to generate more meaningful and semantically coherent chunks. During inference, the framework retrieves information by leveraging both segment-level and cluster-level vector representations, thereby increasing the likelihood of retrieving more precise and contextually relevant information. Evaluations on the NarrativeQA, QuALITY, and QASPER datasets indicate that the proposed method achieved improved results compared to traditional chunking techniques.

Extracting cause and effect phrases from a sentence is an important NLP task, with numerous applications in various domains, including legal, medical, education, and scientific research. There are many unsupervised and supervised methods proposed for solving this task. Among these, unsupervised methods utilize various linguistic tools, including syntactic patterns, dependency tree, dependency relations, etc. among different sentential units for extracting the cause and effect phrases. On the other hand, the contemporary supervised methods use various deep learning based mask language models equipped with a token classification layer for extracting cause and effect phrases. Linguistic tools, specifically, dependency tree, which organizes a sentence into different semantic units have been shown to be very effective for extracting semantic pairs from a sentence, but existing supervised methods do not have any provision for utilizing such tools within their model framework. In this work, we propose DepBERT, which extends a transformer-based model by incorporating dependency tree of a sentence within the model framework. Extensive experiments over three datasets show that DepBERT is better than various state-of-the art supervised causality extraction methods.

Video understanding plays a vital role in bridging low-level visual signals with high-level cognitive reasoning, and is fundamental to applications such as autonomous driving, embodied AI, and the broader pursuit of AGI. The rapid development of large language models (LLMs), particularly those utilizing Chain-of-Thought (CoT) technology, has significantly advanced video reasoning capabilities. However, current approaches primarily depend on textual information for reasoning, overlooking the visual modality in the actual video reasoning process. In contrast, humans naturally re-examine visual content while reasoning. Motivated by this, we introduce a novel video reasoning paradigm: Video-Text Interleaved CoT (ViTCoT), which facilitates more intuitive and cognitively aligned reasoning. To the end, first, we construct the Video-Text Interleaved Benchmark (ViTIB), which is created using MLLMs for key-video selection and manually verified. Furthermore, we extensively explore the potential of the ViTCoT paradigm in the video understanding field. Extensive experiments demonstrate that ViTCoT significantly enhances performance compared to the traditional text-only CoT paradigm and effectively activates more neuron values in MLLMs.

Code large language models (LLMs) enhance programming by understanding and generating code across languages, offering intelligent feedback, bug detection, and code updates through reflection, improving development efficiency and accessibility. While benchmarks (e.g. HumanEval/LiveCodeBench) evaluate code generation and real-world relevance, previous works ignore the scenario of modifying code in repositories. Considering challenges remaining in improving reflection capabilities and avoiding data contamination in dynamic benchmarks, we introduce LiveRepoReflection, a challenging benchmark for evaluating code understanding and generation in multi-file repository contexts, featuring 1,888 rigorously filtered test cases across $6$ programming languages to ensure diversity, correctness, and high difficulty. Further, we create RepoReflection-Instruct, a large-scale, quality-filtered instruction-tuning dataset derived from diverse sources, used to train RepoReflectionCoder through a two-turn dialogue process involving code generation and error-driven repair. The leaderboard evaluates over 40 LLMs to reflect the model performance of repository-based code reflection.

Visually-Rich Document Understanding (VRDU) has emerged as a critical field, driven by the need to automatically process documents containing complex visual, textual, and layout information. Recently, Multimodal Large Language Models (MLLMs) have shown remarkable potential in this domain, leveraging both Optical Character Recognition (OCR)-dependent and OCR-free frameworks to extract and interpret information in document images. This survey reviews recent advancements in MLLM-based VRDU, highlighting three core components: (1) methods for encoding and fusing textual, visual, and layout features; (2) training paradigms, including pretraining strategies, instruction-response tuning, and the trainability of different model modules; and (3) datasets utilized for pretraining, instruction-tuning, and supervised fine-tuning. Finally, we discuss the challenges and opportunities in this evolving field and propose future directions to advance the efficiency, generalizability, and robustness of VRDU systems.

Neurosymbolic artificial intelligence (AI) systems combine neural network and classical symbolic AI mechanisms to exploit the complementary strengths of large-scale, generaliz-able learning and robust, verifiable reasoning. Numerous classifications of neurosymbolic AI illustrate how these two components can be integrated in distinctly different ways. In this work, we propose reinterpreting instruction-tuned large language models as model-grounded symbolic AI systems --where natural language serves as the symbolic layer, and grounding is achieved through the model's internal representation space. Within this framework, we investigate and develop novel learning and reasoning approaches that preserve structural similarities to traditional learning and reasoning paradigms. Preliminary evaluations across axiomatic deductive reasoning procedure of varying complexity provides insights into the effectiveness of our approach towards learning efficiency and reasoning reliability.

An increasing number of NLP applications interact with large language models (LLMs) through black-box APIs, making prompt engineering critical for controlling model outputs. While recent Automatic Prompt Optimization (APO) methods iteratively refine prompts using model-generated feedback, textual gradients, they primarily focus on error correction and neglect valuable insights from correct predictions. This limits both their effectiveness and efficiency. In this paper, we propose a novel APO framework centered on enhancing the feedback mechanism. We reinterpret the textual gradient as a form of negative reinforcement and introduce the complementary positive reinforcement to explicitly preserve beneficial prompt components identified through successful predictions. To mitigate the noise inherent in LLM-generated feedback, we introduce a technique called feedback diversification, which aggregates multiple feedback signals, emphasizing consistent, actionable advice while filtering out outliers. Motivated by the rapid evolution and diversity of available LLMs, we also formalize Continual Prompt Optimization (CPO), addressing the practical challenge of efficiently migrating optimized prompts between different model versions or API providers. Our experiments reveal that naive prompt migration often degrades performance due to loss of critical instructions. In contrast, our approach consistently outperforms strong baselines, achieving significant accuracy improvements, faster convergence, and lower computational costs in both standard and migration scenarios.

LinkedIn, one of the world's largest platforms for professional networking and job seeking, encounters various modeling challenges in building recommendation systems for its job matching product, including cold-start, filter bubbles, and biases affecting candidate-job matching. To address these, we developed the STAR (Signal Integration for Talent And Recruiters) system, leveraging the combined strengths of Large Language Models (LLMs) and Graph Neural Networks (GNNs). LLMs excel at understanding textual data, such as member profiles and job postings, while GNNs capture intricate relationships and mitigate cold-start issues through network effects. STAR integrates diverse signals by uniting LLM and GNN capabilities with industrial-scale paradigms including adaptive sampling and version management. It provides an end-to-end solution for developing and deploying embeddings in large-scale recommender systems. Our key contributions include a robust methodology for building embeddings in industrial applications, a scalable GNN-LLM integration for high-performing recommendations, and practical insights for real-world model deployment.