LLM-based autonomous agents possess capabilities such as reasoning, tool invocation, and environment interaction, enabling the execution of complex multi-step tasks. The internal reasoning process, i.e., thought, of behavioral trajectory significantly influences tool usage and subsequent actions but can introduce potential risks. Even minor deviations in the agent's thought may trigger cascading effects leading to irreversible safety incidents. To address the safety alignment challenges in long-horizon behavioral trajectories, we propose Thought-Aligner, a plug-in dynamic thought correction module. Utilizing a lightweight and resource-efficient model, Thought-Aligner corrects each high-risk thought on the fly before each action execution. The corrected thought is then reintroduced to the agent, ensuring safer subsequent decisions and tool interactions. Importantly, Thought-Aligner modifies only the reasoning phase without altering the underlying agent framework, making it easy to deploy and widely applicable to various agent frameworks. To train the Thought-Aligner model, we construct an instruction dataset across ten representative scenarios and simulate ReAct execution trajectories, generating 5,000 diverse instructions and more than 11,400 safe and unsafe thought pairs. The model is fine-tuned using contrastive learning techniques. Experiments across three agent safety benchmarks involving 12 different LLMs demonstrate that Thought-Aligner raises agent behavioral safety from approximately 50% in the unprotected setting to 90% on average. Additionally, Thought-Aligner maintains response latency below 100ms with minimal resource usage, demonstrating its capability for efficient deployment, broad applicability, and timely responsiveness. This method thus provides a practical dynamic safety solution for the LLM-based agents.

The ReAct (Reasoning + Action) capability in large language models (LLMs) has become the foundation of modern agentic systems. Recent LLMs, such as DeepSeek-R1 and OpenAI o1/o3, exemplify this by emphasizing reasoning through the generation of ample intermediate tokens, which help build a strong premise before producing the final output tokens. In this paper, we introduce Pre-Act, a novel approach that enhances the agent's performance by creating a multi-step execution plan along with the detailed reasoning for the given user input. This plan incrementally incorporates previous steps and tool outputs, refining itself after each step execution until the final response is obtained. Our approach is applicable to both conversational and non-conversational agents. To measure the performance of task-oriented agents comprehensively, we propose a two-level evaluation framework: (1) turn level and (2) end-to-end. Our turn-level evaluation, averaged across five models, shows that our approach, Pre-Act, outperforms ReAct by 70% in Action Recall on the Almita dataset. While this approach is effective for larger models, smaller models crucial for practical applications, where latency and cost are key constraints, often struggle with complex reasoning tasks required for agentic systems. To address this limitation, we fine-tune relatively small models such as Llama 3.1 (8B & 70B) using the proposed Pre-Act approach. Our experiments show that the fine-tuned 70B model outperforms GPT-4, achieving a 69.5% improvement in action accuracy (turn-level) and a 28% improvement in goal completion rate (end-to-end) on the Almita (out-of-domain) dataset.

Ultrasound (US) report generation is a challenging task due to the variability of US images, operator dependence, and the need for standardized text. Unlike X-ray and CT, US imaging lacks consistent datasets, making automation difficult. In this study, we propose a unified framework for multi-organ and multilingual US report generation, integrating fragment-based multilingual training and leveraging the standardized nature of US reports. By aligning modular text fragments with diverse imaging data and curating a bilingual English-Chinese dataset, the method achieves consistent and clinically accurate text generation across organ sites and languages. Fine-tuning with selective unfreezing of the vision transformer (ViT) further improves text-image alignment. Compared to the previous state-of-the-art KMVE method, our approach achieves relative gains of about 2\% in BLEU scores, approximately 3\% in ROUGE-L, and about 15\% in CIDEr, while significantly reducing errors such as missing or incorrect content. By unifying multi-organ and multi-language report generation into a single, scalable framework, this work demonstrates strong potential for real-world clinical workflows.

PDDL-based symbolic task planning remains pivotal for robot autonomy yet struggles with dynamic human-robot collaboration due to scalability, re-planning demands, and delayed plan availability. Although a few neurosymbolic frameworks have previously leveraged LLMs such as GPT-3 to address these challenges, reliance on closed-source, remote models with limited context introduced critical constraints: third-party dependency, inconsistent response times, restricted plan length and complexity, and multi-domain scalability issues. We present Gideon, a novel framework that enables the transition to modern, smaller, local LLMs with extended context length. Gideon integrates a novel problem generator to systematically generate large-scale datasets of realistic domain-problem-plan tuples for any domain, and adapts neurosymbolic planning for local LLMs, enabling on-device execution and extended context for multi-domain support. Preliminary experiments in single-domain scenarios performed on Qwen-2.5 1.5B and trained on 8k-32k samples, demonstrate a valid plan percentage of 66.1% (32k model) and show that the figure can be further scaled through additional data. Multi-domain tests on 16k samples yield an even higher 70.6% planning validity rate, proving extensibility across domains and signaling that data variety can have a positive effect on learning efficiency. Although long-horizon planning and reduced model size make Gideon training much less efficient than baseline models based on larger LLMs, the results are still significant considering that the trained model is about 120x smaller than baseline and that significant advantages can be achieved in inference efficiency, scalability, and multi-domain adaptability, all critical factors in human-robot collaboration. Training inefficiency can be mitigated by Gideon's streamlined data generation pipeline.

Character.ai and other bots such as Chat GPT are based on "large language models" of artificial intelligence. These are trained on vast amounts of data – whether that's websites, articles, books or blog posts - to predict the next word in a sequence. From here, they predict and generate human-like text and interactions. The way mental health chatbots are created varies, but they can be trained in practices such as cognitive behavioural therapy, which helps users to explore how to reframe their thoughts and actions. They can also adapt to the end user's preferences and feedback.

Green was going through a breakup. The reasons for the split itself had been largely unremarkable by breakup standards: Two people, unable to meet each others' needs and struggling to communicate, had decided it was best to part ways. So when Green's ex reached out, unprompted, Green was shocked. The email itself was not notable. Green, a 29-year-old New Yorker, describes it as a typical letter to get after a breakup, an airing of grievances pointing out the ways in which expectations weren't met.

Artificial intelligence can do just as well as humans, if not better, when it comes to persuading others in a debate, and not just because it cannot shout, a study has found. Experts say the results are concerning, not least as it has potential implications for election integrity. "If persuasive AI can be deployed at scale, you can imagine armies of bots microtargeting undecided voters, subtly nudging them with tailored political narratives that feel authentic," said Francesco Salvi, the first author of the research from the Swiss Federal Institute of Technology in Lausanne. He added that such influence was hard to trace, even harder to regulate and nearly impossible to debunk in real time. "I would be surprised if malicious actors hadn't already started to use these tools to their advantage to spread misinformation and unfair propaganda," Salvi said.

Their findings are the latest in a growing body of research demonstrating LLMs' powers of persuasion. The authors warn they show how AI tools can craft sophisticated, persuasive arguments if they have even minimal information about the humans they're interacting with. The research has been published in the journal Nature Human Behavior. "Policymakers and online platforms should seriously consider the threat of coordinated AI-based disinformation campaigns, as we have clearly reached the technological level where it is possible to create a network of LLM-based automated accounts able to strategically nudge public opinion in one direction," says Riccardo Gallotti, an interdisciplinary physicist at Fondazione Bruno Kessler in Italy, who worked on the project. "These bots could be used to disseminate disinformation, and this kind of diffused influence would be very hard to debunk in real time," he says.

You've seen headlines about the market crash and maybe even wondered if now's your shot at finally investing. A stock-picking tool powered by OpenAI is helping regular folks identify strong opportunities with minimal risk. Meet Sterling Stock Picker, the thing that could turn your savings account into an early retirement, extra travel funds, or whatever you wish. Rather than gambling with your hard-earned dollars, this tool helps you research options that match your preferences and risk tolerance, and a lifetime subscription is just 55.19 (reg. Want to dive into the stock market but feel like you're reading a foreign language?

Reinforcement learning from human feedback (RLHF) and, at its core, reward modeling have become a crucial part of training powerful large language models (LLMs). One commonly overlooked factor in training high-quality reward models (RMs) is the effect of the base model, which is becoming more challenging to choose given the rapidly growing pool of LLMs. In this work, we present a systematic analysis of the effect of base model selection on reward modeling performance. Our results show that the performance can be improved by up to 14% compared to the most common (i.e., default) choice. Moreover, we showcase the strong statistical relation between some existing benchmarks and downstream performances. We also demonstrate that the results from a small set of benchmarks could be combined to boost the model selection ($+$18% on average in the top 5-10). Lastly, we illustrate the impact of different post-training steps on the final performance and explore using estimated data distributions to reduce performance prediction error.