Hierarchical Merging is a technique commonly used to summarize very long texts ($>$100K tokens) by breaking down the input into smaller sections, summarizing those sections individually, and then merging or combining those summaries into a final coherent summary. Although it helps address the limitations of large language models (LLMs) with fixed input length constraints, the recursive merging process can amplify LLM hallucinations, increasing the risk of factual inaccuracies. In this paper, we seek to mitigate hallucinations by enriching hierarchical merging with context from the source document. Specifically, we propose different approaches to contextual augmentation ranging from \emph{replacing} intermediate summaries with relevant input context, to \emph{refining} them while using the context as supporting evidence, and \emph{aligning} them implicitly (via citations) to the input. Experimental results on datasets representing legal and narrative domains show that contextual augmentation consistently outperforms zero-shot and hierarchical merging baselines for the Llama 3.1 model family. Our analysis further reveals that refinement methods tend to perform best when paired with extractive summarization for identifying relevant input.

This paper investigates the problem of interactively learning behaviors communicated by a human teacher using positive and negative feedback. Much previous work on this problem has made the assumption that people provide feedback for decisions that is dependent on the behavior they are teaching and is independent from the learner's current policy. We present empirical results that show this assumption to be false -- whether human trainers give a positive or negative feedback for a decision is influenced by the learner's current policy. Based on this insight, we introduce {\em Convergent Actor-Critic by Humans} (COACH), an algorithm for learning from policy-dependent feedback that converges to a local optimum. Finally, we demonstrate that COACH can successfully learn multiple behaviors on a physical robot.

Study authors include Professor Shrikanth Narayanan, Timothy Greer, Dillon Knox, and Benjamin Ma. (Images Courtesy of Narayanan, Greer, Knox, and Ma) Music is an indispensable element in film: it establishes atmosphere and mood, drives the viewer's emotional reactions, and significantly influences the audience's interpretation of the story. In a recent paper published in PLOS One, a research team at the USC Viterbi School of Engineering, led by Professor Shrikanth Narayanan, sought to objectively examine the effect of music on cinematic genres. Their study aimed to determine if AI-based technology could predict the genre of a film based on the soundtrack alone. "By better understanding how music affects the viewer's perception of a film, we gain insights into how film creators can reach their audience in a more compelling way," said Narayanan, University Professor and Niki and Max Nikias Chair in Engineering, professor of electrical and computer engineering and computer science and the director of USC Viterbi's Signal Analysis and Interpretation Laboratory (SAIL). The notion that different film genres are more likely to use certain musical elements in their soundtrack is rather intuitive: a lighthearted romance might include rich string passages and lush, lyrical melodies, while a horror film might instead feature unsettling, piercing frequencies and eerily discordant notes.

Even before the recent George Floyd case, there has been much debate about the extent to which claims of systemic racism are supported by statistical evidence. For example (Ross 2015) claims that unarmed blacks are 3.5 times more likely to be shot by police than unarmed whites when adjusting for relative differences in population size. However, (Fryer 2016) - formally published later as (Fryer 2019) - found that there was no such racial disparity when the data were conditioned on people being stopped by police, and there was a similar conclusion in (Patty and Hanson 2020) that was produced in direct response to public concerns about the Floyd case. In response to Fryer, (Ross, Winterhalder, and McElreath 2018) argued that Fryer's analysis was compromised because it was essentially an example of Simpson's paradox (Simpson 1951; Bickel, Hammel, and O'Connell 1975; Fenton, Neil, and Constantinou 2019) whereby conclusions based on pooled statistics are reversed when drilling down into relevant subcategories. A new paper (Knox, Lowe, and Mummolo 2020) explains why Simpson's paradox is not the only statistical explanation for the apparently contradictory conclusions of Ross and Fryer.

It is with much rejoicing that today we can share that one of our favorite robotics startups, Dash Robotics, is acquiring another of our favorite robotics startups, Bots Alive. Usually, we don't cover acquisitions, or when we do, it's with resigned skepticism--all too often, one company gets completely swallowed by another, and the things that made them unique and exciting simply vanish. The sense that we get from talking with Dash Robotics' CEO Nick Kohut and Bots Alive founder Brad Knox is that the amazing things that Bots Alive does fit right in with the equally amazing but totally different things that Dash Robotics does, and that together, they'll be able to come up with some totally cool (and totally affordable) robotic toys with sophisticated personalities built right in. Part of the reason that we're fans of Dash Robotics and Bots Alive is that they're both successful examples of taking robotics research and turning it directly into a compelling product. Dash Robotics turned UC Berkeley's DASH pop-up hexapod robot into a skittery and blisteringly fast toy called Kamigami that's now being sold in partnership with Mattel for US $50, while Bots Alive's software runs on your phone and gives a $20 Hexbug more brains and personality than an enthusiastic and mildly well trained puppy.

The new chat-bot provides information compiled by doctors for canna-curious patients. AI is all the rage on social platforms these days, from POTUS-based Twitterbots to digital dirty-talkers. On Facebook Messenger, a fresh bundle of algorithms is even offering to give new and experienced cannabis users the medicine-led answers they need fast (if still a bit on the green side, conversationally). ABBI is a new edutainment bot designed to handle all your legal cannabis-related queries, and which is now available for free chatsultations via Messenger. Compliments of the expertly named course hosting and soon-to-be news site WeedHorn, the chatbot is built on OctaneAI's Convos editor, which founder Rick Bakas said he chose to allow the most flexibility.

Any video gamer knows how boring NPCs (non-playable characters) in digital worlds are. Their behavior is simple and predictable and their words entirely scripted by a staff of writers. This makes them uninteresting opponents and unsatisfying companions. We're far more likely to emotionally attach to lifelike characters, like the emo robot sidekicks in the Star Wars franchise, but crafting believable, autonomous entities you can actually interact with is no easy feat. Character models built by artificial intelligence aim to escape the uncanny valley and imbue inanimate objects and digital characters with an aura of realism and life.

Any video gamer knows how boring NPCs (non-playable characters) in digital worlds are. Their behavior is simple and predictable and their words entirely scripted by a staff of writers. This makes them uninteresting opponents and unsatisfying companions. We're far more likely to emotionally attach to lifelike characters, like the emo robot sidekicks in the Star Wars franchise, but crafting believable, autonomous entities you can actually interact with is no easy feat. Character models built by artificial intelligence aim to break out of the uncanny valley and imbue inanimate objects and digital characters with an aura of realism and life.

The artificial intelligence research behind bots_alive promises to usher in lifelike robots at affordable price points. Artificial intelligence is all the rage in robotics these days, and for good reason: Properly implemented, it has the potential to program'bots on the fly. But those toys and others react in predictable ways to changing contexts and situations. One startup, though, purports to have developed an algorithm capable of generating entirely new behaviors dynamically. It's called bots_alive, and it's the brainchild of Brad Knox.

There was no shortage of gadgets at CES, and there will be no shortage at Toy Fair next month, of robots and gadgets promising artificial intelligence -- and generally falling short. But a more modest approach from an actual AI researcher has produced a clever and accessible way to create lifelike behavior through a simple and elegant modification of a popular existing robot. The kit is called bots_alive, and it's looking for a mere $15,000 on Kickstarter. I talked with creator Brad Knox about the tech at CES, and came away pleased with the simplicity of the design at a time when overbearing, talking, dancing robot toys seem to be the norm. It works like this: you start with a Hexbug Spider, a remote control legged toy robot you can buy for $25. They're normally operated with a tiny infrared controller.