Wandering Adventure Party

Researchers showed that bioactive glass administered with gallium oxide in the laboratory can reduce cancer cells and leave noncancerous cells unharmed.

Test-time inference has emerged as a powerful paradigm for enabling language models to ``think'' longer and more carefully about complex challenges, much like skilled human experts. While reinforcement learning (RL) can drive self-improvement in language models on verifiable tasks, some models exhibit substantial gains while others quickly plateau. For instance, we find that Qwen-2.5-3B far exceeds Llama-3.2-3B under identical RL training for the game of Countdown. This discrepancy raises a critical question: what intrinsic properties enable effective self-improvement? We introduce a framework to investigate this question by analyzing four key cognitive behaviors -- verification, backtracking, subgoal setting, and backward chaining -- that both expert human problem solvers and successful language models employ. Our study reveals that Qwen naturally exhibits these reasoning behaviors, whereas Llama initially lacks them. In systematic experimentation with controlled behavioral datasets, we find that priming Llama with examples containing these reasoning behaviors enables substantial improvements during RL, matching or exceeding Qwen's performance. Importantly, the presence of reasoning behaviors, rather than correctness of answers, proves to be the critical factor -- models primed with incorrect solutions containing proper reasoning patterns achieve comparable performance to those trained on correct solutions. Finally, leveraging continued pretraining with OpenWebMath data, filtered to amplify reasoning behaviors, enables the Llama model to match Qwen's self-improvement trajectory. Our findings establish a fundamental relationship between initial reasoning behaviors and the capacity for improvement, explaining why some language models effectively utilize additional computation while others plateau.

Ambition is a predictor of success. But according to a new study, the motives behind it can also lead to lying and cheating.

Lifestyle improvements like adopting a healthy diet or quitting smoking can slow biological aging processes.

Large language models (LLMs) have shown remarkable advancements in chemistry and biomedical research, acting as versatile foundation models for various tasks. We introduce AMP-Designer, an LLM-based approach, for swiftly designing antimicrobial peptides (AMPs) with desired properties. Within 11 days, AMP-Designer achieved the de novo design of 18 AMPs with broad-spectrum activity against Gram-negative bacteria. In vitro validation revealed a 94.4% success rate, with two candidates demonstrating exceptional antibacterial efficacy, minimal hemotoxicity, stability in human plasma, and low potential to induce resistance, as evidenced by significant bacterial load reduction in murine lung infection experiments. The entire process, from design to validation, concluded in 48 days. AMP-Designer excels in creating AMPs targeting specific strains despite limited data availability, with a top candidate displaying a minimum inhibitory concentration of 2.0 micrograms per milliliter against Propionibacterium acnes. Integrating advanced machine learning techniques, AMP-Designer demonstrates remarkable efficiency, paving the way for innovative solutions to antibiotic resistance.

• Researchers found that infectious bacteria in diabetic mice rapidly evolved resistance to antibiotics.
• Controlling blood sugar in the mice via insulin significantly reduced the emergence of drug-resistant bacteria.

• Scientists found that, when deprived of amino acids, cancer cells cooperated to extract and share them from their environment.
• Blocking a protein called CNDP2 shut down this cooperative survival strategy, suggesting a new potential target for cancer treatment.

Experiments in tiny freshwater animals suggest that certain tumors manipulate their host’s body to increase the likelihood of being transmitted to the next generation.

Using social robots is a promising approach for supporting senior citizens in the context of super-aging societies. The essential design factors for achieving socially acceptable robots include effective emotional expressions and cuteness. Past studies have reported the effectiveness of robot-initiated touching behaviors toward interacting partners on these two factors in the context of interaction with adults, although the effects of such touch behaviors on them are unknown in seniors. Therefore, in this study, we investigated the effects of robot-initiated touch behaviors on perceived emotions (valence and arousal) and the feeling of kawaii, a common Japanese adjective for expressing cute, lovely, or adorable. In experiments with Japanese participants (adults: 21–49, seniors: 65–79) using a baby-type robot, our results showed that the robot’s touch significantly increased the perceived valence regardless of the expressed emotions and the ages of the participants. Our results also showed that the robot’s touch was effective in adults in the context of arousal and the feeling of kawaii, but not in seniors. We discussed the differential effects of robot-initiated touch between adults and seniors by focusing on emotional processing in the latter. The findings of this study have implications for designing social robots that have the capability of physical interaction with seniors.

Superconducting magnets inside a fusion reactor will experience conditions that aren’t seen anywhere on Earth. Materials scientists Susie Speller and Chris Grovenor are trying to predict how long these components can last in this extreme environment

Tomato yellow leaf curl virus (TYLCV), a devastating pathogen of tomato crops, is vectored by the whitefly Bemisia tabaci, yet the mechanisms underlying TYLVC epidemics are poorly understood. We found that TYLCV triggers the up-regulation of two β-myrcene biosynthesis genes in tomato, leading to the attraction of nonviruliferous B. tabaci. We also identified BtMEDOR6 as a key whitefly olfactory receptor of β-myrcene involved in the distinct preference of B. tabaci MED for TYLCV-infected plants. TYLCV inhibits the expression of BtMEDOR6, canceling this preference and thereby facilitating TYLCV transmission to uninfected plants. Greenhouse experiments corroborated the role of β-myrcene in whitefly attraction. These findings reveal a sophisticated viral strategy whereby TYLCV modulates both host plant attractiveness and vector olfactory perception to enhance its spread.

The discovery of a biological "third state" between life and death suggests that our cells may possess a form of consciousness. Science is divided.

An experimental treatment for an aggressive and lethal brain cancer has today been published in Nature Medicine, paving the way for a clinical trial to be conducted by researchers at The Brain Cancer Centre.

Would you let a scientist implant a parasite in your brain to treat an illness? The idea of using a parasite as a medical tool might seem outlandish, but it also offers hope for conditions like Parkinsons and Alzheimer’s. Researchers have speculated that if a parasite could deliver drugs directly to the brain, it might help doctors treat these conditions.

An international team of scientists is doing just that. They are using a single-celled parasite known as Toxoplasma gondii, which causes an infection called toxoplasmosis. The parasite naturally travels from the human gut to the central nervous system and delivers proteins to host cells. In the experiment, bioengineers hacked systems within the T. gondii’s cells that produce and release proteins outside the cell, called secretion systems.

As far as animals go, honey bees are world-class dancers.

While not as deep and complex as a Super Bowl half-time show, the bees' moves, known as the “waggle" dance, convey very specific food foraging instructions to their nestmates. The direction the dancer moves explains to other bees which way to go, and the duration of the waggle dance, or the “run,” shows how far to go. Once other bees have been convinced to follow the directions, they are “recruited.” After receiving the instructions, these recruits leave the hive to find the food their sisters were so excited about.