Recent studies have demonstrated the effectiveness of LLM test-time scaling. However, existing approaches to incentivize LLMs' deep thinking abilities generally require large-scale data or significant training efforts. Meanwhile, it remains unclear how to improve the thinking abilities of less powerful base models. In this work, we introduce S$^2$R, an efficient framework that enhances LLM reasoning by teaching models to self-verify and self-correct during inference. Specifically, we first initialize LLMs with iterative self-verification and self-correction behaviors through supervised fine-tuning on carefully curated data. The self-verification and self-correction skills are then further strengthened by both outcome-level and process-level reinforcement learning, with minimized resource requirements, enabling the model to adaptively refine its reasoning process during inference. Our results demonstrate that, with only 3.1k self-verifying and self-correcting behavior initialization samples, Qwen2.5-math-7B achieves an accuracy improvement from 51.0\% to 81.6\%, outperforming models trained on an equivalent amount of long-CoT distilled data. Extensive experiments and analysis based on three base models across both in-domain and out-of-domain benchmarks validate the effectiveness of S$^2$R. Our code and data are available at https://github.com/NineAbyss/S2R.

Titanfall 2: This impressive sci-fi action game for Xbox One can be found for just $3 – at Dollarama stores. While you might pop into your local dollar store to pick up odds and ends – such as school supplies, snacks, or loot bag gifts – be sure to take a walk down the electronics aisle as you'll likely be surprised what you can find. So long as you have reasonable expectations when it comes to the quality and longevity, you won't be disappointed with most of what's available on the shelves. Hey, there are even some stellar games from yesteryear you can buy with spare change. Here's what I found, based on many visits to various discount retailers including Dollar Tree, Dollar General, 99¢ Depot, and Dollarama (in Canada).

Robot workers will soon be roaming the isles of some Schnucks grocery stores in the U.S. The chain announced the introduction of a fleet of robot stock boys named Tally with screens to'make them appear friendly' to its workforce - and while these robo employees don't have limbs and can't physically stock the shelves, they'll be tasked with wandering the isles to check inventory and verify prices. The chain - which has 100 locations in five states - will initially test Tally in two stores before hopefully rolling out the full robot fleet. Tally will start at Schnucks locations in Richmond Heights, Kirkwood, and Town and Country, Missouri. 'This is a big learning experience for us to really understand what the capability is,' Dave Steck, Schnuck Markets' vice president of IT and infrastructure, told the St.Louis Post-Dispatch. Company documents reviewed by Bloomberg before Whole Foods acquisition was announced suggest that automation will be a key strategy used by Amazon.

Watch Google Home, the Internet giant's take on Amazon Echo, as it struggles to answer basic queries posed by Jefferson Graham on #TalkingTech LOS ANGELES - A new look to the Uber app that taps into your contacts to potentially share rides, a deep dive into Google Home, the product that tries to take on Amazon's Echo, but some say fails, subscription based organic dog food, video doorbells and online dollar stores. The extended weekend edition of the Talking Tech podcast weighs in on the week's tech's headlines, and introduces you to some new and very interesting tech startups. The app adds calendar integration--tap it to pick up a ride request for that destination. Uber also taps into your contact list to send push notifications to friend, asks if he/she wants to share ride. Microsoft Teams won't be fully available until the first quarter of 2017.