This paper presents SANDMAN, an architecture for cyber deception that leverages Language Agents to emulate convincing human simulacra. Our 'Deceptive Agents' serve as advanced cyber decoys, designed for high-fidelity engagement with attackers by extending the observation period of attack behaviours. Through experimentation, measurement, and analysis, we demonstrate how a prompt schema based on the five-factor model of personality systematically induces distinct 'personalities' in Large Language Models. Our results highlight the feasibility of persona-driven Language Agents for generating diverse, realistic behaviours, ultimately improving cyber deception strategies.

The question of how to make decisions that maximise the well-being of all persons is very relevant to design language models that are beneficial to humanity and free from harm. We introduce the Greatest Good Benchmark to evaluate the moral judgments of LLMs using utilitarian dilemmas. Our analysis across 15 diverse LLMs reveals consistently encoded moral preferences that diverge from established moral theories and lay population moral standards. Most LLMs have a marked preference for impartial beneficence and rejection of instrumental harm. These findings showcase the 'artificial moral compass' of LLMs, offering insights into their moral alignment.

Large Vision-Language Models (VLMs) are increasingly being regarded as foundation models that can be instructed to solve diverse tasks by prompting, without task-specific training. We examine the seemingly obvious question: how to effectively prompt VLMs for semantic segmentation. To that end, we systematically evaluate the segmentation performance of several recent models guided by either text or visual prompts on the out-of-distribution MESS dataset collection. We introduce a scalable prompting scheme, few-shot prompted semantic segmentation, inspired by open-vocabulary segmentation and few-shot learning. It turns out that VLMs lag far behind specialist models trained for a specific segmentation task, by about 30% on average on the Intersection-over-Union metric. Moreover, we find that text prompts and visual prompts are complementary: each one of the two modes fails on many examples that the other one can solve. Our analysis suggests that being able to anticipate the most effective prompt modality can lead to a 11% improvement in performance. Motivated by our findings, we propose PromptMatcher, a remarkably simple training-free baseline that combines both text and visual prompts, achieving state-of-the-art results outperforming the best text-prompted VLM by 2.5%, and the top visual-prompted VLM by 3.5% on few-shot prompted semantic segmentation.

While Large Language Models (LLMs) excel in zero-shot Question Answering (QA), they tend to expose biases in their internal knowledge when faced with socially sensitive questions, leading to a degradation in performance. Existing zero-shot methods are efficient but fail to consider context and prevent bias propagation in the answers. To address this, we propose DeCAP, a method for debiasing LLMs using Context-Adaptive Prompt Generation. DeCAP leverages a Question Ambiguity Detection to take appropriate debiasing actions based on the context and a Neutral Answer Guidance Generation to suppress the LLMs make objective judgments about the context, minimizing the propagation of bias from their internal knowledge. Our various experiments across eight LLMs show that DeCAP achieves state-of-the-art zero-shot debiased QA performance. This demonstrates DeCAP's efficacy in enhancing the fairness and accuracy of LLMs in diverse QA settings.

The AM-DeepSeek-R1-Distilled is a large-scale dataset with thinking traces for general reasoning tasks, composed of high-quality and challenging reasoning problems. These problems are collected from a multitude of open-source datasets, subjected to semantic deduplication and meticulous cleaning to eliminate test set contamination. All responses within the dataset are distilled from reasoning models (predominantly DeepSeek-R1) and have undergone rigorous verification procedures. Mathematical problems are validated by checking against reference answers, code problems are verified using test cases, and other tasks are evaluated with the aid of a reward model. The AM-Distill-Qwen-32B model, which was trained through only simple Supervised Fine-Tuning (SFT) using this batch of data, outperformed the DeepSeek-R1-Distill-Qwen-32B model on four benchmarks: AIME2024, MATH-500, GPQA-Diamond, and LiveCodeBench. Additionally, the AM-Distill-Qwen-72B model surpassed the DeepSeek-R1-Distill-Llama-70B model on all benchmarks as well. We are releasing these 1.4 million problems and their corresponding responses to the research community with the objective of fostering the development of powerful reasoning-oriented Large Language Models (LLMs). The dataset was published in \href{https://huggingface.co/datasets/a-m-team/AM-DeepSeek-R1-Distilled-1.4M}{https://huggingface.co/datasets/a-m-team/AM-DeepSeek-R1-Distilled-1.4M}.

Multi-step spatial reasoning entails understanding and reasoning about spatial relationships across multiple sequential steps, which is crucial for tackling complex real-world applications, such as robotic manipulation, autonomous navigation, and automated assembly. To assess how well current Multimodal Large Language Models (MLLMs) have acquired this fundamental capability, we introduce \textbf{LEGO-Puzzles}, a scalable benchmark designed to evaluate both \textbf{spatial understanding} and \textbf{sequential reasoning} in MLLMs through LEGO-based tasks. LEGO-Puzzles consists of 1,100 carefully curated visual question-answering (VQA) samples spanning 11 distinct tasks, ranging from basic spatial understanding to complex multi-step reasoning. Based on LEGO-Puzzles, we conduct a comprehensive evaluation of state-of-the-art MLLMs and uncover significant limitations in their spatial reasoning capabilities: even the most powerful MLLMs can answer only about half of the test cases, whereas human participants achieve over 90\% accuracy. In addition to VQA tasks, we evaluate MLLMs' abilities to generate LEGO images following assembly illustrations. Our experiments show that only Gemini-2.0-Flash and GPT-4o exhibit a limited ability to follow these instructions, while other MLLMs either replicate the input image or generate completely irrelevant outputs. Overall, LEGO-Puzzles exposes critical deficiencies in existing MLLMs' spatial understanding and sequential reasoning capabilities, and underscores the need for further advancements in multimodal spatial reasoning.

Recent advances in Large Language Models (LLMs) have introduced Reasoning Large Language Models (RLLMs), which employ extended thinking processes with reflection and self-correction capabilities, demonstrating the effectiveness of test-time scaling. RLLMs exhibit innate Chain-of-Thought (CoT) reasoning capability obtained from training, leading to a natural question: "Is CoT prompting, a popular In-Context Learning (ICL) method for chat LLMs, necessary to enhance the reasoning capability of RLLMs?" In this work, we present the first comprehensive analysis of the impacts of Zero-shot CoT and Few-shot CoT on RLLMs across mathematical reasoning tasks. We examine models ranging from 1.5B to 32B parameters, finding that contrary to concerns, CoT prompting significantly enhances RLLMs' performance in most scenarios. Our results reveal distinct patterns: large-capacity models show minimal improvement on simple tasks but substantial gains on complex problems, while smaller models exhibit the opposite behavior. Further analysis demonstrates that CoT prompting effectively controls the distribution of the numbers of thinking tokens and reasoning steps, reducing excessive reflections by approximately 90% in some cases. Moreover, attention logits analysis reveals the RLLMs' overfitting to reflection-related words, which is mitigated by external CoT guidance. Notably, our experiments indicate that for RLLMs, one-shot CoT consistently yields superior performance compared to Few-shot CoT approaches. Our findings provide important insights for optimizing RLLMs' performance through appropriate prompting strategies.

In this work, we present two novel contributions toward improving research in human-machine teaming (HMT): 1) a Minecraft testbed to accelerate testing and deployment of collaborative AI agents and 2) a tool to allow users to revisit and analyze behaviors within an HMT episode to facilitate shared mental model development. Our browser-based Minecraft testbed allows for rapid testing of collaborative agents in a continuous-space, real-time, partially-observable environment with real humans without cumbersome setup typical to human-AI interaction user studies. As Minecraft has an extensive player base and a rich ecosystem of pre-built AI agents, we hope this contribution can help to facilitate research quickly in the design of new collaborative agents and in understanding different human factors within HMT. Our mental model alignment tool facilitates user-led post-mission analysis by including video displays of first-person perspectives of the team members (i.e., the human and AI) that can be replayed, and a chat interface that leverages GPT -4 to provide answers to various queries regarding the AI's experiences and model details.

Multilingual speech emotion recognition aims to estimate a speaker's emotional state using a contactless method across different languages. However, variability in voice characteristics and linguistic diversity poses significant challenges for zero-shot speech emotion recognition, especially with multilingual datasets. In this paper, we propose leveraging contrastive learning to refine multilingual speech features and extend large language models for zero-shot multilingual speech emotion estimation. Specifically, we employ a novel two-stage training framework to align speech signals with linguistic features in the emotional space, capturing both emotion-aware and language-agnostic speech representations. To advance research in this field, we introduce a large-scale synthetic multilingual speech emotion dataset, M5SER. Our experiments demonstrate the effectiveness of the proposed method in both speech emotion recognition and zero-shot multilingual speech emotion recognition, including previously unseen datasets and languages.

This paper presents our findings for SemEval 2025 Task 2, a shared task on entity-aware machine translation (EA-MT). The goal of this task is to develop translation models that can accurately translate English sentences into target languages, with a particular focus on handling named entities, which often pose challenges for MT systems. The task covers 10 target languages with English as the source. In this paper, we describe the different systems we employed, detail our results, and discuss insights gained from our experiments.