Generative search engines and deep research LLM agents promise trustworthy, source-grounded synthesis, yet users regularly encounter overconfidence, weak sourcing, and confusing citation practices. We introduce DeepTRACE, a novel sociotechnically grounded audit framework that turns prior community-identified failure cases into eight measurable dimensions spanning answer text, sources, and citations. DeepTRACE uses statement-level analysis (decomposition, confidence scoring) and builds citation and factual-support matrices to audit how systems reason with and attribute evidence end-to-end. Using automated extraction pipelines for popular public models (e.g., GPT-4.5/5, You.com, Perplexity, Copilot/Bing, Gemini) and an LLM-judge with validated agreement to human raters, we evaluate both web-search engines and deep-research configurations. Our findings show that generative search engines and deep research agents frequently produce one-sided, highly confident responses on debate queries and include large fractions of statements unsupported by their own listed sources. Deep-research configurations reduce overconfidence and can attain high citation thoroughness, but they remain highly one-sided on debate queries and still exhibit large fractions of unsupported statements, with citation accuracy ranging from 40--80% across systems.

Prompts are crucial for task definition and for improving the performance of large language models (LLM)-based systems. However, existing LLM-based multi-talker (MT) automatic speech recognition (ASR) systems either omit prompts or rely on simple task-definition prompts, with no prior work exploring the design of prompts to enhance performance. In this paper, we propose extracting serialized output prompts (SOP) and explicitly guiding the LLM using structured prompts to improve system performance (SOP-MT-ASR). A Separator and serialized Connectionist Temporal Classification (CTC) layers are inserted after the speech encoder to separate and extract MT content from the mixed speech encoding in a first-speaking-first-out manner. Subsequently, the SOP, which serves as a prompt for LLMs, is obtained by decoding the serialized CTC outputs using greedy search. To train the model effectively, we design a three-stage training strategy, consisting of serialized output training (SOT) fine-tuning, serialized speech information extraction, and SOP-based adaptation. Experimental results on the LibriMix dataset show that, although the LLM-based SOT model performs well in the two-talker scenario, it fails to fully leverage LLMs under more complex conditions, such as the three-talker scenario. The proposed SOP approach significantly improved performance under both two- and three-talker conditions.

Claim decomposition plays a crucial role in the fact-checking process by breaking down complex claims into simpler atomic components and identifying their unfactual elements. Despite its importance, current research primarily focuses on generative methods for decomposition, with insufficient emphasis on evaluating the quality of these decomposed atomic claims. To bridge this gap, we introduce \textbf{DecMetrics}, which comprises three new metrics: \texttt{COMPLETENESS}, \texttt{CORRECTNESS}, and \texttt{SEMANTIC ENTROPY}, designed to automatically assess the quality of claims produced by decomposition models. Utilizing these metrics, we develop a lightweight claim decomposition model, optimizing its performance through the integration of these metrics as a reward function. Through automatic evaluation, our approach aims to set a benchmark for claim decomposition, enhancing both the reliability and effectiveness of fact-checking systems.

Visual Emotion Recognition (VER) is an important research topic due to its wide range of applications, including opinion mining and advertisement design. Extending this capability to recognize emotions at the individual level further broadens its potential applications. Recently, Multimodal Large Language Models (MLLMs) have attracted increasing attention and demonstrated performance comparable to that of conventional VER methods. However, MLLMs are trained on large and diverse datasets containing general opinions, which causes them to favor majority viewpoints and familiar patterns. This tendency limits their performance in a personalized VER, which is crucial for practical and real-world applications, and indicates a key area for improvement. To address this limitation, the proposed method employs discrete prompt tuning inspired by the process of humans' prompt engineering to adapt the VER task to each individual. Our method selects the best natural language representation from the generated prompts and uses it to update the prompt for the realization of accurate personalized VER.

Large language models struggle with rare token prediction, yet the mechanisms driving their specialization remain unclear. Prior work identified specialized ``plateau'' neurons for rare tokens following distinctive three-regime influence patterns \cite{liu2025emergent}, but their functional organization is unknown. We investigate this through neuron influence analyses, graph-based clustering, and attention head ablations in GPT-2 XL and Pythia models. Our findings show that: (1) rare token processing requires additional plateau neurons beyond the power-law regime sufficient for common tokens, forming dual computational regimes; (2) plateau neurons are spatially distributed rather than forming modular clusters; and (3) attention mechanisms exhibit no preferential routing to specialists. These results demonstrate that rare token specialization arises through distributed, training-driven differentiation rather than architectural modularity, preserving context-sensitive flexibility while achieving adaptive capacity allocation.

Recent advances in Large Language Models (LLMs) have been driven by test-time compute scaling - a strategy that improves reasoning by generating longer, sequential thought processes. While effective, this approach encounters a significant bottleneck as computation increases, where further computation offers only marginal performance gains. We argue this ceiling is not an inherent limit of the model's capability but a flaw in the scaling strategy itself, a phenomenon we term "Tunnel Vision", where a model's imperfect initial steps lock it into a suboptimal reasoning path. To overcome this, we introduce a new scaling paradigm: native thought parallelism. We present ParaThinker, an end-to-end framework that trains an LLM to generate multiple, diverse reasoning paths in parallel and synthesize them into a superior final answer. By exploring different lines of thoughts simultaneously, ParaThinker effectively sidesteps the Tunnel Vision issue and unlocks the model's latent reasoning potential. Our approach demonstrates that scaling compute in parallel (width) is a more effective and efficient way to superior reasoning than simply scaling sequentially (depth). On challenging reasoning benchmarks, ParaThinker achieves substantial accuracy improvements over sequential LLMs (12.3% for 1.5B and 7.5% for 7B models on average with 8 parallel paths), while adding only negligible latency overhead (7.1%). This enables smaller models to surpass much larger counterparts and establishes parallel thinking as a critical, efficient dimension for scaling future LLMs.

Test-time scaling has emerged as a powerful paradigm for enhancing the reasoning capabilities of large language models (LLMs) by allocating additional computational resources during inference. However, this paradigm is inherently inefficient due to the generation of redundant and repetitive reasoning traces, leading to significant computational overhead. Speculative decoding offers a promising avenue for mitigating this inefficiency, yet its efficacy in the structured, repetition-rich context of test-time scaling remains largely unexplored. To bridge this gap, we introduce the first comprehensive benchmark designed to evaluate speculative decoding methods for accelerating LLM test-time scaling. Our benchmark provides consistent experimental protocols across representative test-time scaling paradigms (e.g., Best-of-N sampling and multi-round thinking), enabling a fair comparison of three major categories of speculative decoding: model-based, training-based, and n-gram-based methods. Extensive experiments reveal that simple n-gram-based methods effectively capture repetitive patterns, demonstrating unique potential in accelerating test-time scaling. This phenomenon demonstrates the value of integrating n-gram-based methods with model-based or training-based approaches to balance acceleration for both repetitive and diverse reasoning in test-time scaling. We hope this benchmark spurs further research on speculative decoding for test-time scaling, enabling faster and more practical reasoning in LLMs through better handling of repetitive and diverse reasoning paths.

While integrating speech encoder with LLM requires substantial data and resources, use cases face limitations due to insufficient availability. To address this, we propose a solution with a parameter-efficient adapter that converts speech embeddings into LLM-compatible tokens, focusing on end-to-end automatic speech recognition (ASR), named entity recognition (NER), and sentiment analysis (SA). To reduce labeling costs, we employ an LLM-based synthetic dataset annotation technique. The proposed adapter, using 7x fewer trainable parameters, achieves significant performance gains: a 26% relative Word Error Rates (WER) improvement on the LibriSpeech ASR task, a 6.3% relative F1 score increase on the NER task, and a 32% relative F1 score boost on the SA task. Moreover, using advanced techniques such as adding a classifier regularizer and optimizing the LLM with Low-Rank Adaptation (LoRA) yields notable performance gains, with Spoken Language Understanding Evaluation (SLUE) score improvement of 6.6% and 9.5%

We present COCORELI, a hybrid agent framework designed to tackle the limitations of large language models (LLMs) in tasks requiring: following complex instructions, minimizing hallucination, and spatial reasoning. COCORELI integrates medium-sized LLM agents with novel abstraction mechanisms and a discourse module to parse instructions to in-context learn dynamic, high-level representations of the environment. Experiments on natural collaborative construction tasks show that COCORELI outperforms single-LLM CoT and agentic LLM systems, all using larger LLMs. It manages to largely avoid hallucinations, identify missing information, ask for clarifications, and update its learned objects. COCORELI's abstraction abilities extend beyond ENVIRONMENT, as shown in the ToolBench API completion task.

We compare two processing strategies: direct image processing using multi-modal capabilities and a structured parsing approach converting documents to markdown first. Results show native image processing generally outperforms structured approaches, with performance varying across model types and document characteristics. This benchmark provides insights for selecting appropriate models and processing strategies for automated document systems. Our code is available online.