Detecting hidden behaviors in neural networks poses a significant challenge due to minimal prior knowledge and potential adversarial obfuscation. We explore this problem by framing detection as an adversarial game between two teams: the red team trains two similar models, one trained solely on benign data and the other trained on data containing hidden harmful behavior, with the performance of both being nearly indistinguishable on the benign dataset. The blue team, with limited to no information about the harmful behaviour, tries to identify the compromised model. We experiment using CNNs and try various blue team strategies, including Gaussian noise analysis, model diffing, integrated gradients, and adversarial attacks under different levels of hints provided by the red team. Results show high accuracy for adversarial-attack-based methods (100\% correct prediction, using hints), which is very promising, whilst the other techniques yield more varied performance. During our LLM-focused rounds, we find that there are not many parallel methods that we could apply from our study with CNNs. Instead, we find that effective LLM auditing methods require some hints about the undesired distribution, which can then used in standard black-box and open-weight methods to probe the models further and reveal their misalignment. We open-source our auditing games (with the model and data) and hope that our findings contribute to designing better audits.

As Large Language Models (LLMs) become more sophisticated, there is a possibility to harness LLMs to power social media bots. This work investigates the realism of generating LLM-Powered social media bot networks. Through a combination of manual effort, network science and LLMs, we create synthetic bot agent personas, their tweets and their interactions, thereby simulating social media networks. We compare the generated networks against empirical bot/human data, observing that both network and linguistic properties of LLM-Powered Bots differ from Wild Bots/Humans. This has implications towards the detection and effectiveness of LLM-Powered Bots.

Large Language models have demonstrated promising performance in research ideation across scientific domains. Hypothesis development, the process of generating a highly specific declarative statement connecting a research idea with empirical validation, has received relatively less attention. Existing approaches trivially deploy retrieval augmentation and focus only on the quality of the final output ignoring the underlying reasoning process behind ideation. We present $\texttt{HypER}$ ($\textbf{Hyp}$othesis Generation with $\textbf{E}$xplanation and $\textbf{R}$easoning), a small language model (SLM) trained for literature-guided reasoning and evidence-based hypothesis generation. $\texttt{HypER}$ is trained in a multi-task setting to discriminate between valid and invalid scientific reasoning chains in presence of controlled distractions. We find that $\texttt{HypER}$ outperformes the base model, distinguishing valid from invalid reasoning chains (+22\% average absolute F1), generates better evidence-grounded hypotheses (0.327 vs. 0.305 base model) with high feasibility and impact as judged by human experts ($>$3.5 on 5-point Likert scale).

This paper examines how synthetically generated faces and machine learning-based gender classification algorithms are affected by algorithmic lookism, the preferential treatment based on appearance. In experiments with 13,200 synthetically generated faces, we find that: (1) text-to-image (T2I) systems tend to associate facial attractiveness to unrelated positive traits like intelligence and trustworthiness; and (2) gender classification models exhibit higher error rates on "less-attractive" faces, especially among non-White women. These result raise fairness concerns regarding digital identity systems.

Plan-guided summarization attempts to reduce hallucinations in small language models (SLMs) by grounding generated summaries to the source text, typically by targeting fine-grained details such as dates or named entities. In this work, we investigate whether plan-based approaches in SLMs improve summarization in long document, narrative tasks. Narrative texts' length and complexity often mean they are difficult to summarize faithfully. We analyze existing plan-guided solutions targeting fine-grained details, and also propose our own higher-level, narrative-based plan formulation. Our results show that neither approach significantly improves on a baseline without planning in either summary quality or faithfulness. Human evaluation reveals that while plan-guided approaches are often well grounded to their plan, plans are equally likely to contain hallucinations compared to summaries. As a result, the plan-guided summaries are just as unfaithful as those from models without planning. Our work serves as a cautionary tale to plan-guided approaches to summarization, especially for long, complex domains such as narrative texts. Code available at https://github.com/amazon-science/plan-guided-summarization

Although LLM-based agents have attracted significant attention in domains such as software engineering and machine learning research, their role in advancing combinatorial optimization (CO) remains relatively underexplored. This gap underscores the need for a deeper understanding of their potential in tackling structured, constraint-intensive problems -- a pursuit currently limited by the absence of comprehensive benchmarks for systematic investigation. To address this, we introduce CO-Bench, a benchmark suite featuring 36 real-world CO problems drawn from a broad range of domains and complexity levels. CO-Bench includes structured problem formulations and curated data to support rigorous investigation of LLM agents. We evaluate multiple agentic frameworks against established human-designed algorithms, revealing the strengths and limitations of existing LLM agents and identifying promising directions for future research. CO-Bench is publicly available at https://github.com/sunnweiwei/CO-Bench.

Canada's Prime Minister Mark Carney has expressed support for Ukraine's calls for security guarantees as part of any peace deal with Russia, including the possibility of deploying troops to the Eastern European country. During a visit to Kyiv, where he met Ukraine's President Volodymyr Zelenskyy on Sunday, Carney said a group of Ukraine's Western allies, known as the Coalition of the Willing, is working with the United States to bolster Ukrainian defences. "In Canada's judgment, it is not realistic that the only security guarantee could be the strength of the Ukrainian Armed Forces … that needs to be buttressed and reinforced," Carney told reporters. "We are working through – with our allies in Coalition of the Willing and with Ukraine – the modalities of those security guarantees on land, in the air and the sea, and I would not exclude the presence of troops." Three and a half years since Russia's full-scale invasion of Ukraine, US President Donald Trump is leading efforts to end the war.

Turning Point USA founder Charlie Kirk spoke with Fox News Digital about his thoughts of "South Park" parodying him in an upcoming episode, calling it a "badge of honor." "South Park," Comedy Central's gold-standard animated sitcom, has launched its 27th season on America's television screens and, with President Trump back in the White House, politics is back on the menu for creators Trey Parker and Matt Stone. Much like our national media ecosystem, Trump and his presidency are the driving force behind almost every plot line in the first three episodes this year. Much of it is quite funny, but one does wonder: Where was all this hilarious hijinx regarding Joe Biden's "Weekend At Bernie's" presidency? The overarching premise of the season thus far is that, with the election of Trump, wokeness is finally dead.

Russia has accused Ukraine of carrying out a drone attack on a nuclear plant that has caused a fire and damage to an auxiliary transformer as Ukraine celebrates its Independence Day for the 34th time. Sunday's attack forced a 50 percent reduction in the operating capacity at reactor number three at the Kursk Nuclear Power Plant (NPP), close to the border with Ukraine, according to Russian officials, who added that several power and energy facilities were targeted in the overnight strikes. The fire at the nuclear facility was quickly extinguished with no injuries reported, the plant's news service said on Telegram. Two other reactors are operating without power generation, and one is undergoing scheduled repairs, it said, adding that radiation levels were normal. Alexander Khinshtein, the Kursk region's acting governor, said Ukrainian attacks on the plant, 60km (38 miles) from the Russia-Ukraine border, "are a threat to nuclear safety and a violation of all international conventions".

Pete Nicoletti, chief information security officer at Check Point, told Fox News Digital that an FBI-warned scam is now using AI to target seniors. A cybersecurity expert warns that a scam that has been used to drain entire life savings or retirement accounts has become "devastating" for seniors. FBI Los Angeles on July 15 posted a reminder on X about the Phantom Hacker Scam, which has cost Americans over 1 billion since at least 2024, according to the agency. The FBI said the scam targets senior citizens and warns that victims could lose their "life savings." The scam operates in three phases: a "tech support impostor," "financial institution impostor" and a "US government impostor." In the first phase, a tech support impostor will contact victims through text, phone call or email, then direct them to download a program allowing the scammer remote access to their computer.