The same phenomena that let beetles float across ponds and cause Cheerios to cluster together in your cereal bowl can be harnessed to make tiny floating robots. One of these, the Marangoni effect, arises when a fluid with a lower surface tension rapidly spreads out across the surface of a fluid with higher surface tension. This effect is exploited by Stenus beetles, which have evolved to zip across ponds by secreting a substance called stenusin, as well as soap-powered toy boats. To investigate how this could be used by engineers, Jackson Wilt at Harvard University and his colleagues 3D-printed round, plastic pucks around a centimetre in diameter. Inside each was an air chamber for buoyancy and a tiny fuel tank containing alcohol, which has a lower surface tension than water, in concentrations from 10 to 50 per cent.

While large language models (LLMs) have shown impressive capabilities across a wide range of domains, they still encounter significant challenges in reasoning tasks that require gathering evidence over multiple turns and drawing logical conclusions from this evidence. These challenges present significant obstacles for LLM chat user interfaces, which rely on multi-turn interactions to facilitate effective collaboration. This limitation leads to real-world issues; for example, service chatbots must gather necessary information from customers over multiple turns to diagnose and resolve problems effectively. Despite the multi-turn nature of many real-world LLM use cases, most existing benchmarks rely on carefully curated single-turn tests, which often blur the line between memorization and genuine reasoning. To address this, we introduce the Wason Inductive Logic Test (WILT), a simple yet challenging multi-turn reasoning benchmark designed to resist memorization. WILT is inspired by the Wason 2-4-6 task (Wason, 1960), where participants must infer a basic boolean function involving three variables (e.g., x < y < z) by proposing test cases (such as (2, 4, 6)). In WILT, each test starts from a clean slate, with only the initial instructions provided, preventing models from relying on pre-learned responses. Over several turns, models must interact with the environment by suggesting test cases to narrow the possible hypotheses and ultimately infer the hidden function based on the outcomes. Our findings reveal that LLMs struggle with this task, exhibiting distinct strengths and weaknesses: some are better at narrowing down the hypothesis space by proposing valuable test cases, while others are more adept at deducing the hidden function from observed cases. Despite these variations, the best-performing model achieves only 28% accuracy, highlighting a significant gap in LLM performance on complex multi-turn reasoning tasks. Large language models (LLMs) powered by the transformer architecture (Vaswani, 2017) have enabled a new computing paradigm driven by natural language. These models are increasingly integrated into day-to-day life beyond the machine learning research space, where they help many people with common tasks. These models interact with users through multi-turn conversations, a capability of next-token-prediction models bolstered via instruction-tuning (Mishra et al., 2021) and alignment post-training phases (Ouyang et al., 2022).

When an optimal solution is not required, satisficing search methods such as greedy best-first search are often used to find solutions quickly. In work on satisficing search, there has been substantial attention devoted to how to solve problems associated with local minima or plateaus in the heuristic function. One technique that has been shown to be quite promising is using an alternative heuristic function that does not estimate cost-to-go, but rather estimates distance-to-go. There is currently little beyond intuition to explain its superiority. We begin by empirically showing that the success of the distance-to-go heuristic appears related to its having smaller local minima. We then discuss a reasonable theoretical model of heuristics and show that, under this model, the expected size of local minima is higher for a cost-to-go heuristic than a distance-to-go heuristic, offering a possible explanation as to why distance-to-go heuristics tend to outperform cost-to-go heuristics.

Artificial intelligence-powered tools are rapidly becoming more accessible, including for people in the more remote corners of the globe. This is good news for smallholder farmers, who can use handheld technologies to run their farms more efficiently, linking them to markets, extension workers, satellite images, and climate information. The technology is also becoming a first line of defense against crop diseases and pests that can potentially destroy their harvests. A new smartphone tool developed for banana farmers scans plants for signs of five major diseases and one common pest. In testing in Colombia, the Democratic Republic of the Congo, India, Benin, China, and Uganda, the tool provided a 90 percent successful detection rate. This work is a step towards creating a satellite-powered, globally connected network to control disease and pest outbreaks, say the researchers who developed the technology.