In real practice, questions are typically complex and knowledge-intensive, requiring Large Language Models (LLMs) to recognize the multifaceted nature of the question and reason across multiple information sources. Iterative and adaptive retrieval, where LLMs decide when and what to retrieve based on their reasoning, has been shown to be a promising approach to resolve complex, knowledge-intensive questions. However, the performance of such retrieval frameworks is limited by the accumulation of reasoning errors and misaligned retrieval results. To overcome these limitations, we propose TreeRare (Syntax Tree-Guided Retrieval and Reasoning), a framework that utilizes syntax trees to guide information retrieval and reasoning for question answering. Following the principle of compositionality, TreeRare traverses the syntax tree in a bottom-up fashion, and in each node, it generates subcomponent-based queries and retrieves relevant passages to resolve localized uncertainty. A subcomponent question answering module then synthesizes these passages into concise, context-aware evidence. Finally, TreeRare aggregates the evidence across the tree to form a final answer. Experiments across five question answering datasets involving ambiguous or multi-hop reasoning demonstrate that TreeRare achieves substantial improvements over existing state-of-the-art methods.

Fox News Flash top entertainment and celebrity headlines are here. James Cameron's stance on artificial intelligence has evolved over the past few years, and he feels Hollywood needs to embrace it in a few different ways. Cameron joined the board of directors for Stability AI last year, explaining his decision on the "Boz to the Future" podcast last week. "The goal was to understand the space, to understand what's on the minds of the developers," he said. How much resources you have to throw at it to create a new model that does a purpose-built thing, and my goal was to try to integrate it into a VFX workflow." He continued by saying the shift to AI is a necessary one. James Cameron wants Hollywood to implement AI more for big-budget films. WHAT IS ARTIFICIAL INTELLIGENCE (AI)? If we want to continue to see the kinds of movies that I've always loved and that I like to make and that I will go to see – 'Dune,' 'Dune: Part Two' or one of my films or big effects-heavy, CG-heavy films – we've got to figure out how to cut the cost of that in half. That's about doubling their speed to completion on a given shot, so your cadence is faster and your throughput cycle is faster, and artists get to move on and do other cool things and then other cool things, right? Cameron doesn't think films are ultimately "a big target" for companies like OpenAI. "Their goal is not to make GenAI movies.

Pretraining Large Language Models (LLMs) on large corpora of textual data is now a standard paradigm. When using these LLMs for many downstream applications, it is common to additionally bake in new knowledge (e.g., time-critical news, or private domain knowledge) into the pretrained model either through RAG-based-prompting, or fine-tuning. However, the optimal methodology for the model to gain such new knowledge remains an open question. In this paper, we present Retrieval Augmented FineTuning (RAFT), a training recipe that improves the model's ability to answer questions in a "open-book" in-domain settings. In RAFT, given a question, and a set of retrieved documents, we train the model to ignore those documents that don't help in answering the question, which we call, distractor documents. RAFT accomplishes this by citing verbatim the right sequence from the relevant document that would help answer the question. This coupled with RAFT's chain-of-thought-style response helps improve the model's ability to reason. In domain-specific RAG, RAFT consistently improves the model's performance across PubMed, HotpotQA, and Gorilla datasets, presenting a post-training recipe to improve pre-trained LLMs to in-domain RAG. RAFT's code and demo are open-sourced at github.com/ShishirPatil/gorilla.

We posit that large language models (LLMs) should be capable of expressing their intrinsic uncertainty in natural language. For example, if the LLM is equally likely to output two contradicting answers to the same question, then its generated response should reflect this uncertainty by hedging its answer (e.g., "I'm not sure, but I think..."). We formalize faithful response uncertainty based on the gap between the model's intrinsic confidence in the assertions it makes and the decisiveness by which they are conveyed. This example-level metric reliably indicates whether the model reflects its uncertainty, as it penalizes both excessive and insufficient hedging. We evaluate a variety of aligned LLMs at faithfully communicating uncertainty on several knowledge-intensive question answering tasks. Our results provide strong evidence that modern LLMs are poor at faithfully conveying their uncertainty, and that better alignment is necessary to improve their trustworthiness.

By leveraging the retrieval of information from external knowledge databases, Large Language Models (LLMs) exhibit enhanced capabilities for accomplishing many knowledge-intensive tasks. However, due to the inherent flaws of current retrieval systems, there might exist irrelevant information within those retrieving top-ranked passages. In this work, we present a comprehensive investigation into the robustness of LLMs to different types of irrelevant information under various conditions. We initially introduce a framework to construct high-quality irrelevant information that ranges from semantically unrelated, partially related, and related to questions. Furthermore, our analysis demonstrates that the constructed irrelevant information not only scores highly on similarity metrics, being highly retrieved by existing systems, but also bears semantic connections to the context. Our investigation reveals that current LLMs still face challenges in discriminating highly semantically related information and can be easily distracted by these irrelevant yet misleading contents. Besides, we also find that current solutions for handling irrelevant information have limitations in improving the robustness of LLMs to such distractions.

One of the longest labor crises in Hollywood history is finally coming to an end. SAG-AFTRA, the union representing tens of thousands of actors, reached a tentative deal for a new contract with entertainment companies on Wednesday, clearing the way for the $134 billion American movie and television business to swing back into motion. Hollywood's assembly lines have been at a near-standstill since May because of a pair of strikes by writers and actors, resulting in financial pain for studios and for many of the two million Americans -- makeup artists, set builders, location scouts, chauffeurs, casting directors -- who work in jobs directly or indirectly related to making TV shows and films. Upset about streaming-service pay and fearful of fast-developing artificial intelligence technology, actors joined screenwriters on picket lines in July. The writers had walked out in May over similar concerns.

My union, the Writers Guild of America (WGA), has been on strike since May 2, when our contract with the studios – the Alliance of Motion Picture and Television Producers (AMPTP) – expired after talks broke down. So, I woke up that morning and headed over to Disney Studios in Burbank, California, wearing a pair of Mickey Mouse ears my niece had cajoled me into buying on a trip to Disneyland. I covered the ears with two words: "FAIR" and "CONTRACT". I've been doing much the same thing for more than 100 days since then. A few of our demands: increased wages, commensurate with industry growth and inflation; writers should get a proportionate cut if shows do well; that cut should grow if more people watch my show in an increasing number of countries; weekly pay for screenwriters; restrictions on the use of artificial intelligence and minimum staffing for writers' rooms.

My name is Simon Rich and I'm a screenwriter. I've never written an opinion piece before. I've always preferred to speak through my fictional characters, because they're played by actors who are better looking. But I happen to be childhood friends with a scientist from OpenAI, and some of the stuff he's shown me is so messed up that I felt the need to write this article. I hope you will take a few minutes to read it while picturing me as Paul Rudd. When most people think about artificial intelligence, they think about ChatGPT.

Veritone CEO Ryan Steelberg says the Writers Guild of America needs to make sure their writers are protected as AI becomes more popular. Nearly two weeks into the national writers' strike spearheaded by the Writers Guild of America (WGA), little progress has been made between both sides. The WGA has a litany of requests for the Alliance of Motion Picture and Television Producers (AMPTP). Per its website, the WGA has specific proposals with regard to artificial intelligence, including the "regulation of AI on minimum basic agreement (MBA) -covered projects; AI can't write or rewrite literary material; can't be used as source material; and MBA-covered material can't be used to train AI." When it comes to these provisions that surround artificial intelligence, studios have put the kibosh on writers' requests, instead suggesting annual meetings to review evolving technology.

One of the more harrowing reads for writers concerned about artificial intelligence encroaching on their livelihoods is a study commissioned by OpenAI itself. Published in March, it places writers in the "fully exposed" category. This means that, according to OpenAI, a large language model (LLM) could reduce the time it takes for them to carry out their work by at least 50 percent. AI can already score in the 93rd percentile on SAT reading exams; it can already produce bad stories and poems. Directors are discussing the possibilities of AI-generated scripts.