One can also improve LLMs' performance on a specific task by

Large language models (LLMs) solve problems more accurately and interpretably when instructed to work out the answer step by step using a ``chain-of-thought'' (CoT) prompt. One can also improve LLMs' performance on a specific task by supervised fine-tuning, i.e., by using gradient ascent on some tunable parameters to maximize the average log-likelihood of correct answers from a labeled training set. Naively combining CoT with supervised tuning requires supervision not just of the correct answers, but also of detailed rationales that lead to those answers; these rationales are expensive to produce by hand. Instead, we propose a fine-tuning strategy that tries to maximize the \emph{marginal} log-likelihood of generating a correct answer using CoT prompting, approximately averaging over all possible rationales. The core challenge is sampling from the posterior over rationales conditioned on the correct answer; we address it using a simple Markov-chain Monte Carlo (MCMC) expectation-maximization (EM) algorithm inspired by the self-taught reasoner (STaR), memoized wake-sleep, Markovian score climbing, and persistent contrastive divergence. This algorithm also admits a novel control-variate technique that drives the variance of our gradient estimates to zero as the model improves. Applying our technique to GSM8K and the tasks in BIG-Bench Hard, we find that this MCMC-EM fine-tuning technique typically improves the model's accuracy on held-out examples more than STaR or prompt-tuning with or without CoT.

Compute-in-Memory (CiM), built upon non-volatile memory (NVM) devices, is promising for accelerating deep neural networks (DNNs) owing to its in-situ data processing capability and superior energy efficiency. Unfortunately, the well-trained model parameters, after being mapped to NVM devices, can often exhibit large deviations from their intended values due to device variations, resulting in notable performance degradation in these CiM-based DNN accelerators. There exists a long list of solutions to address this issue. However, they mainly focus on improving the mean performance of CiM DNN accelerators. How to guarantee the worst-case performance under the impact of device variations, which is crucial for many safety-critical applications such as self-driving cars, has been far less explored. In this work, we propose to use the k-th percentile performance (KPP) to capture the realistic worst-case performance of DNN models executing on CiM accelerators. Through a formal analysis of the properties of KPP and the noise injection-based DNN training, we demonstrate that injecting a novel right-censored Gaussian noise, as opposed to the conventional Gaussian noise, significantly improves the KPP of DNNs. We further propose an automated method to determine the optimal hyperparameters for injecting this right-censored Gaussian noise during the training process. Our method achieves up to a 26% improvement in KPP compared to the state-of-the-art methods employed to enhance DNN robustness under the impact of device variations.

When a person joins an online-dating app, and then starts texting some of the people they've met on that app, and then makes plans to hang out with some of those people in the hopes of making out, they have a reasonable, limited expectation of privacy. Hardly anyone expects what happened to the mythological figure of "West Elm Caleb," a bumbling villain of the New York dating scene and hapless victim of the internet. In January, after a couple of New York women with substantial TikTok followings discovered that they had been dating Caleb simultaneously, it quickly came out that he was guilty of other crimes--sending the same Spotify playlist to multiple people, for instance, and not returning text messages. One woman recalled how he had told her that he found it harder to go on dates in the winter, because of the cold. Pretty soon brands were getting in on the West Elm Caleb conversation, as finding any excuse to talk about this pretty average dater in New York City became engagement-metric gold.

Commercial buildings are voracious consumers of energy. The buildings and building construction sectors contribute around 30% of global energy consumption and almost 40% of CO2 emissions (direct and indirect), according to the IEA. Sustainable buildings are more than trendy, they're key to reducing humans' carbon footprint on the planet. Building owners deploy smart building solutions and building-management systems to not only cut operations costs and comply with regulations but also to do their part in reducing carbon emissions without sacrificing occupant comfort in the process. So how do you make an already smart, efficient building smarter and more efficient?

Fox News Flash top headlines for May 6 are here. Check out what's clicking on Foxnews.com A Kentucky man has been charged with murder for fatally punching his 1-year-old son in the head after becoming angry over losing a video game, authorities said Sunday. Anthony Trice, 26, was watching the baby Friday when he grew enraged over losing the game, threw his controller and struck the infant in the head, the Louisville Metro Police Department said. Trice tried to comfort the baby, carrying him into the kitchen, but dropped him, Louisville station WAVE-TV reported.

Thousands of people showed up Wednesday for a chance to pack and ship products to Amazon customers, as the e-commerce company held a giant job fair at nearly a dozen U.S. warehouses. Although the wages offered will make it hard for some to make ends meet, many of the candidates were excited by the prospect of health insurance and other benefits, as well as advancement opportunities. It's common for Amazon to ramp up its shipping center staff in August to prepare for holiday shopping. But the magnitude of its current hiring spree underscores Amazon's growth when traditional retailers are closing stores -- and blaming Amazon for a shift to buying goods online. Amazon said it received "a record-breaking 20,000 applications" and hired thousands of people on the spot, and will hire more in the coming days.