Breakthroughs, discoveries, and DIY tips sent every weekday. Rubbing a black snail on a wart and impailing the creature with a thorn will make the bumps go away. Giving a donkey some bread will treat whooping cough . Mumps can be cured if you rub your head on the back of a pig . They may sound a bit strange now, but folk remedies like these are an important part of human history.

Can language models (LMs) self-refine their own responses? This question is increasingly relevant as a wide range of real-world user interactions involve refinement requests. However, prior studies have largely tested LMs' refinement abilities on verifiable tasks such as competition math or symbolic reasoning with simplified scaffolds, whereas users often pose open-ended queries and provide varying degrees of feedback on what they desire. The recent advent of reasoning models that exhibit self-reflection patterns in their chains-of-thought further motivates this question. To analyze this, we introduce RefineBench, a benchmark of 1,000 challenging problems across 11 domains paired with a checklist-based evaluation framework. We evaluate two refinement modes: (1) guided refinement, where an LM is provided natural language feedback, and (2) self-refinement, where LMs attempt to improve without guidance. In the self-refinement setting, even frontier LMs such as Gemini 2.5 Pro and GPT-5 achieve modest baseline scores of 31.3% and 29.1%, respectively, and most models fail to consistently improve across iterations (e.g., Gemini-2.5-Pro gains only +1.8%, while DeepSeek-R1 declines by -0.1%). By contrast, in guided refinement, both proprietary LMs and large open-weight LMs (>70B) can leverage targeted feedback to refine responses to near-perfect levels within five turns. These findings suggest that frontier LMs require breakthroughs to self-refine their incorrect responses, and that RefineBench provides a valuable testbed for tracking progress.

After deflecting the US Department of Justice's attack on its illegal monopoly in online search, Google is facing another attempt to dismantle its internet empire in a trial focused on abusive tactics in digital advertising. The trial that opened Monday in an Alexandria, Virginia, federal court revolves around the harmful conduct that resulted in US district Judge Leonie Brinkema declaring parts of Google's digital advertising technology to be an illegal monopoly in April. The judge found that Google has been engaging in behavior that stifles competition to the detriment of online publishers that depend on the system for revenue. Google and the justice department will spend the next two weeks in court presenting evidence in a "remedy" trial that will culminate in Brinkema issuing a ruling on how to restore fair market conditions. If the justice department gets its way, Brinkema will order Google to sell parts of its ad technology - a proposal that the company's lawyers warned would "invite disruption and damage" to consumers and the internet's ecosystem.

Google will not be forced to sell its Chrome browser, a federal judge ruled on Tuesday in the tech giant's ongoing legal battle over being ruled a monopoly last year. The company will be barred from certain exclusive deals with device makers and must share data from its search engine with competitors, the judge ruled. Judge Amit Mehta's ruling follows months of speculation surrounding what penalties Google would face as a result of his decision last year that the company violated antitrust laws as it built what he called an online search monopoly. The ruling, one of the most significant antitrust cases in decades, resulted in an additional hearing in April to determine what actions the government should take as a remedy. Mehta's decision to allow Google to keep Chrome represents a more lenient outcome for the company than what federal prosecutors requested: force the tech giant sell off its marquee search product and to ban it from entering the browser market for five years.

Perplexity AI said it has made a 34.5bn unsolicited all-cash offer for Alphabet's Google Chrome browser. The deal, if Alphabet agreed to it, would also require financing above the startup's most recently reported valuation of 18bn. The nearly three-year-old startup's purchase of Chrome, if approved, would give the company access to its more than three billion users as regulatory pressure weighs on Google's control over the tech industry. Perplexity did not disclose on Tuesday how it plans to fund the offer, but has raised 1bn in funding from investors including SoftBank and the semiconductor chip giant Nvidia. Several funds have said they would finance the deal in full if Alphabet accepts, the Reuters news agency reported citing unnamed sources familiar with the matter.

Privacy-Preserving Federated Learning (PPFL) is a decentralized machine learning approach where multiple clients train a model collaboratively. PPFL preserves the privacy and security of a client's data without exchanging it. However, ensuring that data at each client is of high quality and ready for federated learning (FL) is a challenge due to restricted data access. In this paper, we introduce CADRE (Customizable Assurance of Data Readiness) for federated learning (FL), a novel framework that allows users to define custom data readiness (DR) metrics, rules, and remedies tailored to specific FL tasks. CADRE generates comprehensive DR reports based on the user-defined metrics, rules, and remedies to ensure datasets are prepared for FL while preserving privacy. We demonstrate a practical application of CADRE by integrating it into an existing PPFL framework. We conducted experiments across six datasets and addressed seven different DR issues. The results illustrate the versatility and effectiveness of CADRE in ensuring DR across various dimensions, including data quality, privacy, and fairness. This approach enhances the performance and reliability of FL models as well as utilizes valuable resources.

Centuries-old Islamic medical texts like Avicenna's Canon of Medicine and the Prophetic Tibb-e-Nabawi encode a wealth of preventive care, nutrition, and holistic therapies, yet remain inaccessible to many and underutilized in modern AI systems. Existing language-model benchmarks focus narrowly on factual recall or user preference, leaving a gap in validating culturally grounded medical guidance at scale. We propose a unified evaluation pipeline, Tibbe-AG, that aligns 30 carefully curated Prophetic-medicine questions with human-verified remedies and compares three LLMs (LLaMA-3, Mistral-7B, Qwen2-7B) under three configurations: direct generation, retrieval-augmented generation, and a scientific self-critique filter. Each answer is then assessed by a secondary LLM serving as an agentic judge, yielding a single 3C3H quality score. Retrieval improves factual accuracy by 13%, while the agentic prompt adds another 10% improvement through deeper mechanistic insight and safety considerations. Our results demonstrate that blending classical Islamic texts with retrieval and self-evaluation enables reliable, culturally sensitive medical question-answering.

Google has been back in federal court to fend off the United States Department of Justice's attempt to topple its internet empire at the same time it is navigating a pivotal shift to artificial intelligence (AI) that could undercut its power. On Friday, the legal and technological threats facing Google were among the key issues being dissected during the closing arguments of a legal proceeding that will determine the changes imposed upon the company in the wake of its dominant search engine being declared an illegal monopoly by US District Judge Amit Mehta last year. Brandishing evidence presented during a recent three-week stretch of hearings, Justice Department lawyers are attempting to persuade Mehta to order a radical shake-up that includes a ban on Google paying to lock its search engine in as the default on smart devices and an order requiring the company to sell its Chrome browser. Google lawyers say only minor concessions are needed, especially as the upheaval triggered by advances in artificial intelligence already are reshaping the search landscape, as alternative, conversational search options are rolling out from AI startups that are hoping to use the Department of Justice's four-and-half-year-old case to gain the upper hand in the next technological frontier. Mehta used Friday's hearing to ask probing and pointed questions to lawyers for both sides while hinting that he was seeking a middle ground between the two camps' proposed remedies.

A key challenge in MT evaluation is the inherent noise and inconsistency of human ratings. Regression-based neural metrics struggle with this noise, while prompting LLMs shows promise at system-level evaluation but performs poorly at segment level. In this work, we propose ReMedy, a novel MT metric framework that reformulates translation evaluation as a reward modeling task. Instead of regressing on imperfect human ratings directly, ReMedy learns relative translation quality using pairwise preference data, resulting in a more reliable evaluation. In extensive experiments across WMT22-24 shared tasks (39 language pairs, 111 MT systems), ReMedy achieves state-of-the-art performance at both segment- and system-level evaluation. Specifically, ReMedy-9B surpasses larger WMT winners and massive closed LLMs such as MetricX-13B, XCOMET-Ensemble, GEMBA-GPT-4, PaLM-540B, and finetuned PaLM2. Further analyses demonstrate that ReMedy delivers superior capability in detecting translation errors and evaluating low-quality translations.

The Justice Department said in a filing that Google will have to break up its network of myriad, overlapping businesses and services, upholding the previous administration's proposal. The DOJ reiterated Google will have to sell the Chrome browser -- saying, last year, that selling off Chrome "will permanently stop Google's control of this critical search access point and allow rival search engines the ability to access the browser that for many users is a gateway to the internet." Google is likely to file its own alternate remedies, of course. In a December filing, the company said the Justice Department's original remedies went "overboard" and reflected an "interventionist agenda." But Google is huge, and the DOJ is trying to grasp how its parts intermingle and make it less monopolistic.