The rise of large language models (LLMs) has significantly influenced the quality of information in decision-making systems, leading to the prevalence of AI-generated content and challenges in detecting misinformation and managing conflicting information, or "inter-evidence conflicts." This study introduces a method for generating diverse, validated evidence conflicts to simulate real-world misinformation scenarios. We evaluate conflict detection methods, including Natural Language Inference (NLI) models, factual consistency (FC) models, and LLMs, on these conflicts (RQ1) and analyze LLMs' conflict resolution behaviors (RQ2). Our key findings include: (1) NLI and LLM models exhibit high precision in detecting answer conflicts, though weaker models suffer from low recall; (2) FC models struggle with lexically similar answer conflicts, while NLI and LLM models handle these better; and (3) stronger models like GPT-4 show robust performance, especially with nuanced conflicts. For conflict resolution, LLMs often favor one piece of conflicting evidence without justification and rely on internal knowledge if they have prior beliefs.

This work explores sequential model editing in large language models (LLMs), a critical task that involves modifying internal knowledge within LLMs continuously through multi-round editing, each incorporating updates or corrections to adjust the model outputs without the need for costly retraining. Existing model editing methods, especially those that alter model parameters, typically focus on single-round editing and often face significant challenges in sequential model editing-most notably issues of model forgetting and failure. To address these challenges, we introduce a new model editing method, namely \textbf{N}euron-level \textbf{S}equential \textbf{E}diting (NSE), tailored for supporting sequential model editing. Specifically, we optimize the target layer's hidden states using the model's original weights to prevent model failure. Furthermore, we iteratively select neurons in multiple layers for editing based on their activation values to mitigate model forgetting. Our empirical experiments demonstrate that NSE significantly outperforms current modifying parameters model editing methods, marking a substantial advancement in the field of sequential model editing. Our code is released on \url{https://github.com/jianghoucheng/NSE}.

Language models require tokenized inputs. However, tokenization strategies for continuous data like audio and vision are often based on simple heuristics such as fixed sized convolutions or discrete clustering, which do not necessarily align with the semantic structure of the data. For speech in particular, the high resolution of waveforms (16,000 samples/second or more) presents a significant challenge as speech-based language models have had to use several times more tokens per word than text-based language models. In this work, we introduce a controllable self-supervised technique to merge speech representations into coarser syllable-like units while still preserving semantic information. We do this by 1) extracting noisy boundaries through analyzing correlations in pretrained encoder losses and 2) iteratively improving model representations with a novel distillation technique. Our method produces controllable-rate semantic units at as low as 5Hz and 60bps and achieves SotA in syllabic segmentation and clustering. Using these coarse tokens, we successfully train SyllableLM, a Speech Language Model (SpeechLM) that matches or outperforms current SotA SpeechLMs on a range of spoken language modeling tasks. SyllableLM also achieves significant improvements in efficiency with a 30x reduction in training compute and a 4x wall-clock inference speedup. Learning to generate speech solely from listening to spoken language is a fundamental task in speech processing. It requires abstracting beyond the underlying acoustics of speech into phones, syllables, words, and sentences to process correlations across long ranges of time. But while current textual language models (Touvron et al., 2023; Zhang et al., 2022; Brown et al., 2020) can compose highly realistic text, language models on spoken language still struggle to output semantically meaningful speech.

The SS matrices are downsampled to 5 5. The results indicate that, compared to MusicGen, our method produces The new wave of generative models has been explored in the samples that more closely resemble the Pond5 samples literature to generate music. Jukebox [1] is based on Hierarchical in terms of long-term temporal consistency and the diversity VQ-VAEs [2] to generate multiple minutes of music. of recurring sections. Jukebox is one of the earliest purely learning-based models that could generate longer than one minute of music with some degree of structural coherence. Notably, the authors mention that the generated music at a small scale of multiple learn musical structures and forms at all scales. However, seconds is coherent, and at a larger scale, beyond one minute, none of the models in the literature has demonstrated musical it lacks musical form.

Direct Preference Optimization (DPO) has emerged as a stable, scalable, and efficient solution for language model alignment. Despite its empirical success, the $\textit{optimization}$ properties, particularly the impact of samplers on its convergence rates, remain underexplored. In this paper, we provide a rigorous analysis of DPO's $\textit{convergence rates}$ with different sampling strategies under the exact gradient setting, revealing a surprising separation: uniform sampling achieves $\textit{linear}$ convergence, while our proposed online sampler achieves $\textit{quadratic}$ convergence. We further adapt the sampler to practical settings by incorporating posterior distributions and $\textit{logit mixing}$, demonstrating significant improvements over previous approaches. On Safe-RLHF dataset, our method exhibits a $4.5$% improvement over vanilla DPO and a $3.0$% improvement over on-policy DPO; on Iterative-Prompt, our approach outperforms vanilla DPO, on-policy DPO, and Hybrid GSHF by over $4.2$%. Our results not only offer insights into the theoretical standing of DPO but also pave the way for potential algorithm designs in the future.

Meta, the owner of Facebook and Instagram, announced on Friday it had built a new artificial intelligence model called Movie Gen that can create realistic-seeming video and audio clips in response to user prompts, claiming it can rival tools from leading media generation startups like OpenAI and ElevenLabs. Samples of Movie Gen's creations provided by Meta showed videos of animals swimming and surfing, as well as clips using people's real photos to depict them performing actions like painting on a canvas. Movie Gen also can generate background music and sound effects synced to the content of the videos, Meta said in a blogpost. Users can also edit existing videos with the model. In one such video, Meta had the tool insert pompoms into the hands of a man running by himself in the desert, while in another it changed a parking lot on which a man was skateboarding from dry ground into one covered by a splashing puddle.

Third-year student Jack Quinlan was confident he knew what I was going to ask before we conducted our interview. He wasn't psychic, and I hadn't fed him questions – he'd just done a trial run on ChatGPT. Prior to our meeting, the software engineering and neuroscience undergraduate logged on to the program to generate the kinds of questions a "professional journalist at the Guardian" would ask a student about artificial intelligence at universities. "What prompted your university to begin using generative AI tools in education?" the software version of me began. "How have students and educators at your university responded to the introduction of generative AI? Have there been any challenges and concerns raised?"

Meta just announced its own media-focused AI model, called Movie Gen, that can be used to generate realistic video and audioclips. The company shared multiple 10-second clips generated with Movie Gen, including a Moo Deng-esque baby hippo swimming around, to demonstrate its capabilities. While the tool is not yet available for use, this Movie Gen announcement comes shortly after its Meta Connect event, which showcased new and refreshed hardware and the latest version of its large language model, Llama 3.2. Going beyond the generation of straightforward text-to-video clips, the Movie Gen model can make targeted edits to an existing clip, like adding an object into someone's hands or changing the appearance of a surface. In one of the example videos from Meta, a woman wearing a VR headset was transformed to look like she was wearing steampunk binoculars.

This week, Microsoft started rolling out the Windows 11 2024 update, but it quickly became clear that the company was far more eager to unveil new features for its Copilot AI and Copilot AI PCs. In this episode, Devindra and Cherlynn chat about Microsoft's current AI priorities, and what it means for people with older PCs. Also, we discuss the death of HoloLens and Microsoft giving up on AR as Meta, Apple and even Snap build for an augmented reality future. Listen below or subscribe on your podcast app of choice. If you've got suggestions or topics you'd like covered on the show, be sure to email us or drop a note in the comments! And be sure to check out our other podcast, Engadget News! Tech debt led to Sonos' disastrous app relaunch, will they be able to win users back? Google is making Gmail summaries more useful and adding a "happening soon" tab to your inbox – 41:11 Harvard students hack together facial recognition for Meta's smart glasses that instantly doxes strangers – 44:00 ...

When used as a surrogate objective for maximum likelihood estimation in latent variable models, the evidence lower bound (ELBO) produces state-of-the-art results. Inspired by this, we consider the extension of the ELBO to a family of lower bounds defined by a particle filter's estimator of the marginal likelihood, the filtering variational objectives (FIVOs). FIVOs take the same arguments as the ELBO, but can exploit a model's sequential structure to form tighter bounds. We present results that relate the tightness of FIVO's bound to the variance of the particle filter's estimator by considering the generic case of bounds defined as log-transformed likelihood estimators. Experimentally, we show that training with FIVO results in substantial improvements over training the same model architecture with the ELBO on sequential data.