Large Language Models (LLMs) have shown promise in character imitation, enabling immersive and engaging conversations. However, they often generate content that is irrelevant or inconsistent with a character's background. We attribute these failures to: (1) the inability to accurately recall character-specific knowledge due to entity ambiguity, and (2) a lack of awareness of the character's cognitive boundaries. To address these issues, we propose RoleRAG, a retrieval-based framework that integrates efficient entity disambiguation for knowledge indexing with a boundary-aware retriever for extracting contextually appropriate information from a structured knowledge graph. Experiments on role-playing benchmarks show that RoleRAG's calibrated retrieval helps both general-purpose and role-specific LLMs better align with character knowledge and reduce hallucinated responses.

Role-playing systems powered by large language models (LLMs) have become increasingly influential in emotional communication applications. However, these systems are susceptible to character hallucinations, where the model deviates from predefined character roles and generates responses that are inconsistent with the intended persona. This paper presents the first systematic analysis of character hallucination from an attack perspective, introducing the RoleBreak framework. Our framework identifies two core mechanisms-query sparsity and role-query conflict-as key factors driving character hallucination. Leveraging these insights, we construct a novel dataset, RoleBreakEval, to evaluate existing hallucination mitigation techniques. Our experiments reveal that even enhanced models trained to minimize hallucination remain vulnerable to attacks. To address these vulnerabilities, we propose a novel defence strategy, the Narrator Mode, which generates supplemental context through narration to mitigate role-query conflicts and improve query generalization. Experimental results demonstrate that Narrator Mode significantly outperforms traditional refusal-based strategies by reducing hallucinations, enhancing fidelity to character roles and queries, and improving overall narrative coherence.

Large language models (LLMs) can be used to serve as agents to simulate human behaviors, given the powerful ability to understand human instructions and provide high-quality generated texts. Such ability stimulates us to wonder whether LLMs can simulate a person in a higher form than simple human behaviors. Therefore, we aim to train an agent with the profile, experience, and emotional states of a specific person instead of using limited prompts to instruct ChatGPT API. In this work, we introduce Character-LLM that teach LLMs to act as specific people such as Beethoven, Queen Cleopatra, Julius Caesar, etc. Our method focuses on editing profiles as experiences of a certain character and training models to be personal simulacra with these experiences. To assess the effectiveness of our approach, we build a test playground that interviews trained agents and evaluates whether the agents \textit{memorize} their characters and experiences. Experimental results show interesting observations that help build future simulacra of humankind.

There are two aspects of machine learning and artificial intelligence: (1) interpreting information, and (2) inventing new useful information. Much advance has been made for (1) with a focus on pattern recognition techniques (e.g., interpreting visual data). This paper focuses on (2) with intelligent duplication (ID) for invention. We explore the possibility of learning a specific individual's creative reasoning in order to leverage the learned expertise and talent to invent new information. More specifically, we employ a deep learning system to learn from the great composer Beethoven and capture his composition ability in a hash-based knowledge base. This new form of knowledge base provides a reasoning facility to drive the music composition through a novel music generation method.

The research explores the steerability of Large Language Models (LLMs), particularly OpenAI's ChatGPT iterations. By employing a behavioral psychology framework called OCEAN (Openness, Conscientiousness, Extroversion, Agreeableness, Neuroticism), we quantitatively gauged the model's responsiveness to tailored prompts. When asked to generate text mimicking an extroverted personality, OCEAN scored the language alignment to that behavioral trait. In our analysis, while "openness" presented linguistic ambiguity, "conscientiousness" and "neuroticism" were distinctly evoked in the OCEAN framework, with "extroversion" and "agreeableness" showcasing a notable overlap yet distinct separation from other traits. Our findings underscore GPT's versatility and ability to discern and adapt to nuanced instructions. Furthermore, historical figure simulations highlighted the LLM's capacity to internalize and project instructible personas, precisely replicating their philosophies and dialogic styles. However, the rapid advancements in LLM capabilities and the opaque nature of some training techniques make metric proposals degrade rapidly. Our research emphasizes a quantitative role to describe steerability in LLMs, presenting both its promise and areas for further refinement in aligning its progress to human intentions.

Academics from the University of Cambridge have revealed that they are on the hunt for'super memorisers'. These are people with exceptional memories, and are wanted to take part in a study which could uncover why some are much better at remembering than others. But it may not just be down to natural born ability, as there are some things you can do that have been scientifically proven to help improve your memory. As well as doing brain teasers, there are some less conventional ways, including eating chocolate, walking backwards and spending time in the sunshine. MailOnline takes a look at the strangest techniques scientists have discovered that could turn you into a super memoriser.

On Wednesday morning, I saw a tweet from games magazine PC Gamer that made me leak from the eyes with laughter. It contained a video, in which a wide-eyed, pained-looking cartoon trombonist struggled to hit the notes of Beethoven's Fifth while the composer himself stared sombrely out of the screen in evident disapproval. It is a golden comedic combination of terrible music, fart noises, earnestness and absurdity. This is the video game Trombone Champ, and it has since gone wildly viral. I've been playing rhythm games for more than 20 years, from Beatmania to Guitar Hero to Amplitude via fun musical contraptions in Japanese arcades, and I take them embarrassingly seriously.

Composing a symphony involves many parts to harmonize and rules to follow. When the Beethoven project began in 2019, "Most AI available at the time couldn't continue an uncompleted piece of music beyond a few additional seconds," Elgammal explained in an article for The Conversation. Fortunately, Beethoven left more than 50 sketches behind that alluded to a complete picture of this symphony. Though the sketches can serve as excellent input for the AI, they are fragmentary and almost indecipherable due to his idiosyncratic handwriting. To truly capture the essence of Beethoven's composition, the team also brought on composers, musicologists, and musical historians, intending to teach the AI "both Beethoven's entire body of work and his creative process," Elgammal writes.

Music is liquid architecture; architecture is frozen music.--Attributed to Goethe But Missa solemnis 2.0, a collaboration between pioneering media artist and director Refik Anadol and The Philadelphia Orchestra (April 7, 9, 10, supported by The Pew Center for Arts & Heritage), brings Goethe's pithy saying to stunning visual and sonic life in ways the German literary giant never could have imagined. Beethoven completed his Missa solemnis in 1823. Despite being regarded as one of his most stunning musical creations, the piece is rarely performed. The composer's partner in this century-spanning project, Refik Anadol, was born in Istanbul. In 2008, while still an undergrad there, he presented his first digital art installation.

For this performance of Beethoven with the Philadelphia Orchestra, Anadol "taught" his system by feeding it hundreds of thousands of images of Renaissance art and architecture, the kind of aesthetics that Beethoven himself would have been consuming when he wrote his celebrated works. "We are taking from the Renaissance era every single building ever done, every single sculpture ever created, and every single painting ever done," Anadol said. "These are amazingly large cultural data. We are trying to make an AI to dream these beautiful cultural elements of humanity." The system uses fluid dynamics algorithms to generate animation effects resembling flowing water and wind through hair.