Large Language Models (LLMs) are vulnerable to adversarial prompt based injects. These injects could jailbreak or exploit vulnerabilities within these models with explicit prompt requests leading to undesired responses. In the context of investigating prompt injects, the challenge is the sheer volume of input prompts involved that are likely to be largely benign. This investigative challenge is further complicated by the semantics and subjectivity of the input prompts involved in the LLM conversation with its user and the context of the environment to which the conversation is being carried out. Hence, the challenge for AI security investigators would be two-fold. The first is to identify adversarial prompt injects and then to assess whether the input prompt is contextually benign or adversarial. For the first step, this could be done using existing AI security solutions like guardrails to detect and protect the LLMs. Guardrails have been developed using a variety of approaches. A popular approach is to use signature based. Another popular approach to develop AI models to classify such prompts include the use of NLP based models like a language model. However, in the context of conducting an AI security investigation of prompt injects, these guardrails lack the ability to aid investigators in triaging or assessing the identified input prompts. In this applied research exploration, we explore the use of a text generation capabilities of LLM to detect prompt injects and generate explanation for its detections to aid AI security investigators in assessing and triaging of such prompt inject detections. The practical benefit of such a tool is to ease the task of conducting investigation into prompt injects.

Compared to traditional machine learning models, recent large language models (LLMs) can exhibit multi-task-solving capabilities through multiple dialogues and multi-modal data sources. These unique characteristics of LLMs, beyond their large size, make their deployment more challenging during the inference stage. Specifically, (i) deploying LLMs on local devices faces computational, memory, and energy resource issues, while (ii) deploying them in the cloud cannot guarantee real-time service and incurs communication/usage costs. In this paper, we design a local-cloud LLM inference offloading (LCIO) system, featuring (i) a large-scale cloud LLM that can handle multi-modal data sources and (ii) a lightweight local LLM that can process simple tasks at high speed. LCIO employs resource-constrained reinforcement learning (RCRL) to determine where to make the inference (i.e., local vs. cloud) and which multi-modal data sources to use for each dialogue/task, aiming to maximize the long-term reward (which incorporates response quality, latency, and usage cost) while adhering to resource constraints. We also propose M4A1, a new dataset that accounts for multi-modal, multi-task, multi-dialogue, and multi-LLM characteristics, to investigate the capabilities of LLMs in various practical scenarios. We demonstrate the effectiveness of LCIO compared to baselines, showing significant savings in latency and cost while achieving satisfactory response quality.

Counterfactual reasoning is widely recognized as one of the most challenging and intricate aspects of causality in artificial intelligence. In this paper, we evaluate the performance of large language models (LLMs) in counterfactual reasoning. In contrast to previous studies that primarily focus on commonsense causal reasoning, where LLMs often rely on prior knowledge for inference, we specifically assess their ability to perform counterfactual inference using a set of formal rules. To support this evaluation, we introduce a new benchmark dataset, CounterBench, comprising 1K counterfactual reasoning questions. The dataset is designed with varying levels of difficulty, diverse causal graph structures, distinct types of counterfactual questions, and multiple nonsensical name variants. Our experiments demonstrate that counterfactual reasoning poses a significant challenge for LLMs, with most models performing at levels comparable to random guessing. To enhance LLM's counterfactual reasoning ability, we propose a novel reasoning paradigm, CoIn, which guides LLMs through iterative reasoning and backtracking to systematically explore counterfactual solutions. Experimental results show that our method significantly improves LLM performance on counterfactual reasoning tasks and consistently enhances performance across different LLMs.Our dataset is available at https://huggingface.co/datasets/CounterBench/CounterBench.

Speculative decoding accelerates inference in large language models (LLMs) by generating multiple draft tokens simultaneously. However, existing methods often struggle with token misalignment between the training and decoding phases, limiting their performance. To address this, we propose GRIFFIN, a novel framework that incorporates a token-alignable training strategy and a token-alignable draft model to mitigate misalignment. The training strategy employs a loss masking mechanism to exclude highly misaligned tokens during training, preventing them from negatively impacting the draft model's optimization. The token-alignable draft model introduces input tokens to correct inconsistencies in generated features. Experiments on LLaMA-series and Vicuna models demonstrate that GRIFFIN achieves an average acceptance length improvement of over 7\% and a speedup ratio exceeding 8%, outperforming current SoTAs as shown in Fig. 1 (a) and (b).

Catastrophic forgetting (CF) poses a significant challenge in machine learning, where a model forgets previously learned information upon learning new tasks. Despite the advanced capabilities of Large Language Models (LLMs), they continue to face challenges with CF during continual learning. The majority of existing research focuses on analyzing forgetting patterns through a singular training sequence, thereby overlooking the intricate effects that diverse tasks have on model behavior. Our study explores CF across various settings, discovering that model forgetting is influenced by both the specific training tasks and the models themselves. To this end, we interpret forgetting by examining the function vector (FV), a compact representation of functions in LLMs, offering a model-dependent indicator for the occurrence of CF. Through theoretical and empirical analyses, we demonstrated that CF in LLMs primarily stems from biases in function activation rather than the overwriting of task processing functions. Leveraging these insights, we propose a novel function vector guided training methodology, incorporating a regularization technique to stabilize the FV and mitigate forgetting. Empirical tests on four benchmarks confirm the effectiveness of our proposed training method, substantiating our theoretical framework concerning CF and model function dynamics. We plan to make our code publicly accessible in the near future.

Being able to thoroughly assess massive multi-task language understanding (MMLU) capabilities is essential for advancing the applicability of multilingual language models. However, preparing such benchmarks in high quality native language is often costly and therefore limits the representativeness of evaluation datasets. While recent efforts focused on building more inclusive MMLU benchmarks, these are conventionally built using machine translation from high-resource languages, which may introduce errors and fail to account for the linguistic and cultural intricacies of the target languages. In this paper, we address the lack of native language MMLU benchmark especially in the under-represented Turkic language family with distinct morphosyntactic and cultural characteristics. We propose two benchmarks for Turkic language MMLU: TUMLU is a comprehensive, multilingual, and natively developed language understanding benchmark specifically designed for Turkic languages. It consists of middle- and high-school level questions spanning 11 academic subjects in Azerbaijani, Crimean Tatar, Karakalpak, Kazakh, Tatar, Turkish, Uyghur, and Uzbek. We also present TUMLU-mini, a more concise, balanced, and manually verified subset of the dataset. Using this dataset, we systematically evaluate a diverse range of open and proprietary multilingual large language models (LLMs), including Claude, Gemini, GPT, and LLaMA, offering an in-depth analysis of their performance across different languages, subjects, and alphabets. To promote further research and development in multilingual language understanding, we release TUMLU-mini and all corresponding evaluation scripts.

Deploying large language models (LLMs) in edge-cloud environments requires an efficient routing strategy to balance cost and response quality. Traditional approaches prioritize either human-preference data or accuracy metrics from benchmark datasets as routing criteria, but these methods suffer from rigidity and subjectivity. Moreover, existing routing frameworks primarily focus on accuracy and cost, neglecting response quality from a human preference perspective. In this work, we propose the Confidence-Driven LLM Router, a novel framework that leverages uncertainty estimation to optimize routing decisions. To comprehensively assess routing performance, we evaluate both system cost efficiency and response quality. In particular, we introduce the novel use of LLM-as-a-Judge to simulate human rating preferences, providing the first systematic assessment of response quality across different routing strategies. Extensive experiments on MT-Bench, GSM8K, and MMLU demonstrate that our approach outperforms state-of-the-art routing methods, achieving superior response quality while maintaining cost efficiency.

Natural language interfaces for NoSQL databases are increasingly vital in the big data era, enabling users to interact with complex, unstructured data without deep technical expertise. However, most recent advancements focus on English, leaving a gap for multilingual support. This paper introduces MultiTEND, the first and largest multilingual benchmark for natural language to NoSQL query generation, covering six languages: English, German, French, Russian, Japanese and Mandarin Chinese. Using MultiTEND, we analyze challenges in translating natural language to NoSQL queries across diverse linguistic structures, including lexical and syntactic differences. Experiments show that performance accuracy in both English and non-English settings remains relatively low, with a 4%-6% gap across scenarios like fine-tuned SLM, zero-shot LLM, and RAG for LLM. To address the aforementioned challenges, we introduce MultiLink, a novel framework that bridges the multilingual input to NoSQL query generation gap through a Parallel Linking Process. It breaks down the task into multiple steps, integrating parallel multilingual processing, Chain-of-Thought (CoT) reasoning, and Retrieval-Augmented Generation (RAG) to tackle lexical and structural challenges inherent in multilingual NoSQL generation. MultiLink shows enhancements in all metrics for every language against the top baseline, boosting execution accuracy by about 15% for English and averaging a 10% improvement for non-English languages.

While increasing training compute has significantly improved the performance of large language models (LLMs), similar gains have not been observed when scaling inference compute. We hypothesize that the primary issue lies in the uniformity of LLM outputs, which leads to inefficient sampling as models repeatedly generate similar but inaccurate responses. Motivated by an intriguing relationship between solution accuracy (Pass@10) and response diversity, we propose DivSampling-a novel and versatile sampling technique designed to enhance the diversity of candidate solutions by introducing prompt perturbations.DivSampling incorporates two categories of perturbations: task-agnostic approaches, which are general and not tailored to any specific task, and task-specific approaches, which are customized based on task content. Our theoretical analysis demonstrates that, under mild assumptions, the error rates of responses generated from diverse prompts are significantly lower compared to those produced by stationary prompts. Comprehensive evaluations across various tasks -including reasoning, mathematics, and code generation - highlight the effectiveness of DivSampling in improving solution accuracy. This scalable and efficient approach offers a new perspective on optimizing test-time inference, addressing limitations in current sampling strategies.

Large Language Models (LLMs) demonstrate impressive performance across diverse tasks, yet confidence calibration remains a challenge. Miscalibration - where models are overconfident or underconfident - poses risks, particularly in high-stakes applications. This paper presents an empirical study on LLM calibration, examining how model size, distractors, and question types affect confidence alignment. We introduce an evaluation framework to measure overconfidence and investigate whether multiple-choice formats mitigate or worsen miscalibration. Our findings show that while larger models (e.g., GPT-4o) are better calibrated overall, they are more prone to distraction, whereas smaller models benefit more from answer choices but struggle with uncertainty estimation. Unlike prior work, which primarily reports miscalibration trends, we provide actionable insights into failure modes and conditions that worsen overconfidence. These findings highlight the need for calibration-aware interventions and improved uncertainty estimation methods.