Uncategorized

Earlier this week, MIT Technology Review published its annual list of Ten Breakthrough Technologies. As always, it features technologies that made the news last year, and which—for better or worse—stand to make waves in the coming years. They’re the technologies you should really be paying attention to. This year’s list includes tech that’s set to transform the energy industry, artificial intelligence, space travel—and of course biotech and health. Our breakthrough biotechnologies for 2026 involve editing a baby’s genes and, separately, resurrecting genes from ancient species. We also included a controversial technology that offers parents the chance to screen their embryos for characteristics like height and intelligence. Here’s the story behind our biotech choices. A base-edited baby! In August 2024, KJ Muldoon was born with a rare genetic disorder that allowed toxic ammonia to build up in his blood. The disease can be fatal, and KJ was at risk of developing neurological disorders. At the time, his best bet for survival involved waiting for a liver transplant. Then he was offered an experimental gene therapy—a personalized “base editing” treatment designed to correct the specific genetic “misspellings” responsible for his disease. It seems to have worked! Three doses later, KJ is doing well. He took his first steps in December, shortly before spending his first Christmas at home. KJ’s story is hugely encouraging. The team behind his treatment is planning a clinical trial for infants with similar disorders caused by different genetic mutations. The team members hope to win regulatory approval on the back of a small trial—a move that could make the expensive treatment (KJ’s cost around $1 million) more accessible, potentially within a few years. Others are getting in on the action, too. Fyodor Urnov, a gene-editing scientist at the University of California, Berkeley, assisted the team that developed KJ’s treatment. He recently cofounded Aurora Therapeutics, a startup that hopes to develop gene-editing drugs for another disorder called phenylketonuria (PKU). The goal is to obtain regulatory approval for a single drug that can then be adjusted or personalized for individuals without having to go through more clinical trials. US regulators seem to be amenable to the idea and have described a potential approval pathway for such “bespoke, personalized therapies.” Watch this space. Gene resurrection It was a big year for Colossal Biosciences, the biotech company hoping to “de-extinct” animals like the woolly mammoth and the dodo. In March, the company created what it called “woolly mice”—rodents with furry coats and curly whiskers akin to those of woolly mammoths. The company made an even more dramatic claim the following month, when it announced it had created three dire wolves. These striking snow-white animals were created by making 20 genetic changes to the DNA of gray wolves based on genetic research on ancient dire wolf bones, the company said at the time. Whether these animals can really be called dire wolves is debatable, to say the least. But the technology behind their creation is undeniably fascinating. We’re talking about the extraction and analysis of ancient DNA, which can then be introduced into cells from other, modern-day species. Analysis of ancient DNA can reveal all sorts of fascinating insights into human ancestors and other animals. And cloning, another genetic tool used here, has applications not only in attempts to re-create dead pets but also in wildlife conservation efforts. Read more here. Embryo scoring IVF involves creating embryos in a lab and, typically, “scoring” them on their likelihood of successful growth before they are transferred to a person’s uterus. So far, so uncontroversial. Recently, embryo scoring has evolved. Labs can pinch off a couple of cells from an embryo, look at its DNA, and screen for some genetic diseases. That list of diseases is increasing. And now some companies are taking things even further, offering prospective parents the opportunity to select embryos for features like height, eye color, and even IQ. This is controversial for lots of reasons. For a start, there are many, many factors that contribute to complex traits like IQ (a score that doesn’t capture all aspects of intelligence at any rate). We don’t have a perfect understanding of those factors, or how selecting for one trait might influence another. Some critics warn of eugenics. And others note that whichever embryo you end up choosing, you can’t control exactly how your baby will turn out (and why should you?!). Still, that hasn’t stopped Nucleus, one of the companies offering these services, from inviting potential customers to have their “best baby.” Read more here. This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

arXiv:2601.09667v1 Announce Type: cross Abstract: Multi-agent systems have evolved into practical LLM-driven collaborators for many applications, gaining robustness from diversity and cross-checking. However, multi-agent RL (MARL) training is resource-intensive and unstable: co-adapting teammates induce non-stationarity, and rewards are often sparse and high-variance. Therefore, we introduce textbf{Multi-Agent Test-Time Reinforcement Learning (MATTRL)}, a framework that injects structured textual experience into multi-agent deliberation at inference time. MATTRL forms a multi-expert team of specialists for multi-turn discussions, retrieves and integrates test-time experiences, and reaches consensus for final decision-making. We also study credit assignment for constructing a turn-level experience pool, then reinjecting it into the dialogue. Across challenging benchmarks in medicine, math, and education, MATTRL improves accuracy by an average of 3.67% over a multi-agent baseline, and by 8.67% over comparable single-agent baselines. Ablation studies examine different credit-assignment schemes and provide a detailed comparison of how they affect training outcomes. MATTRL offers a stable, effective and efficient path to distribution-shift-robust multi-agent reasoning without tuning.

arXiv:2601.09633v1 Announce Type: new Abstract: Taxonomies form the backbone of structured knowledge representation across diverse domains, enabling applications such as e-commerce catalogs, semantic search, and biomedical discovery. Yet, manual taxonomy expansion is labor-intensive and cannot keep pace with the emergence of new concepts. Existing automated methods rely on point-based vector embeddings, which model symmetric similarity and thus struggle with the asymmetric “is-a” relationships that are fundamental to taxonomies. Box embeddings offer a promising alternative by enabling containment and disjointness, but they face key issues: (i) unstable gradients at the intersection boundaries, (ii) no notion of semantic uncertainty, and (iii) limited capacity to represent polysemy or ambiguity. We address these shortcomings with TaxoBell, a Gaussian box embedding framework that translates between box geometries and multivariate Gaussian distributions, where means encode semantic location and covariances encode uncertainty. Energy-based optimization yields stable optimization, robust modeling of ambiguous concepts, and interpretable hierarchical reasoning. Extensive experimentation on five benchmark datasets demonstrates that TaxoBell significantly outperforms eight state-of-the-art taxonomy expansion baselines by 19% in MRR and around 25% in Recall@k. We further demonstrate the advantages and pitfalls of TaxoBell with error analysis and ablation studies.

arXiv:2404.03471v5 Announce Type: replace Abstract: Bias in large language models (LLMs) has many forms, from overt discrimination to implicit stereotypes. Counterfactual bias evaluation is a widely used approach to quantifying bias and often relies on template-based probes that explicitly state group membership. It aims to measure whether the outcome of a task performed by an LLM is invariant to a change in group membership. In this work, we find that template-based probes can introduce systematic distortions in bias measurements. Specifically, we consistently find that such probes suggest that LLMs classify text associated with White race as negative at disproportionately elevated rates. This is observed consistently across a large collection of LLMs, over several diverse template-based probes, and with different classification approaches. We hypothesize that this arises artificially due to linguistic asymmetries present in LLM pretraining data, in the form of markedness, (e.g., Black president vs. president) and templates used for bias measurement (e.g., Black president vs. White president). These findings highlight the need for more rigorous methodologies in counterfactual bias evaluation, ensuring that observed disparities reflect genuine biases rather than artifacts of linguistic conventions.

arXiv:2601.09233v1 Announce Type: cross Abstract: The prevailing post-training paradigm for Large Reasoning Models (LRMs)–Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL)–suffers from an intrinsic optimization mismatch: the rigid supervision inherent in SFT induces distributional collapse, thereby exhausting the exploration space necessary for subsequent RL. In this paper, we reformulate SFT within a unified post-training framework and propose Gibbs Initialization with Finite Temperature (GIFT). We characterize standard SFT as a degenerate zero-temperature limit that suppresses base priors. Conversely, GIFT incorporates supervision as a finite-temperature energy potential, establishing a distributional bridge that ensures objective consistency throughout the post-training pipeline. Our experiments demonstrate that GIFT significantly outperforms standard SFT and other competitive baselines when utilized for RL initialization, providing a mathematically principled pathway toward achieving global optimality in post-training. Our code is available at https://github.com/zzy1127/GIFT.

arXiv:2601.08747v1 Announce Type: new Abstract: Current context augmentation methods, such as retrieval-augmented generation, are essential for solving knowledge-intensive reasoning tasks.However, they typically adhere to a rigid, brute-force strategy that executes retrieval at every step. This indiscriminate approach not only incurs unnecessary computational costs but also degrades performance by saturating the context with irrelevant noise. To address these limitations, we introduce Agentic Context Evolution (ACE), a framework inspired by human metacognition that dynamically determines whether to seek new evidence or reason with existing knowledge. ACE employs a central orchestrator agent to make decisions strategically via majority voting.It aims to alternate between activating a retriever agent for external retrieval and a reasoner agent for internal analysis and refinement. By eliminating redundant retrieval steps, ACE maintains a concise and evolved context. Extensive experiments on challenging multi-hop QA benchmarks demonstrate that ACE significantly outperforms competitive baselines in accuracy while achieving efficient token consumption.Our work provides valuable insights into advancing context-evolved generation for complex, knowledge-intensive tasks.

arXiv:2601.08750v1 Announce Type: new Abstract: Recent developments in natural language processing highlight text as an emerging data source for ecology. Textual resources carry unique information that can be used in complementarity with geospatial data sources, thus providing insights at the local scale into environmental conditions and properties hidden from more traditional data sources. Leveraging textual information in a spatial context presents several challenges. First, the contribution of textual data remains poorly defined in an ecological context, and it is unclear for which tasks it should be incorporated. Unlike ubiquitous satellite imagery or environmental covariates, the availability of textual data is sparse and irregular; its integration with geospatial data is not straightforward. In response to these challenges, this work proposes an attention-based approach that combines aerial imagery and geolocated text within a spatial neighbourhood, i.e. integrating contributions from several nearby observations. Our approach combines vision and text representations with a geolocation encoding, with an attention-based module that dynamically selects spatial neighbours that are useful for predictive tasks.The proposed approach is applied to the EcoWikiRS dataset, which combines high-resolution aerial imagery with sentences extracted from Wikipedia describing local environmental conditions across Switzerland. Our model is evaluated on the task of predicting 103 environmental variables from the SWECO25 data cube. Our approach consistently outperforms single-location or unimodal, i.e. image-only or text-only, baselines. When analysing variables by thematic groups, results show a significant improvement in performance for climatic, edaphic, population and land use/land cover variables, underscoring the benefit of including the spatial context when combining text and image data.

arXiv:2601.08160v1 Announce Type: new Abstract: Agentic memory systems have become critical for enabling LLM agents to maintain long-term context and retrieve relevant information efficiently. However, existing memory frameworks suffer from a fundamental limitation: they perform exhaustive retrieval across the entire storage layer regardless of query characteristics. This brute-force approach creates severe latency bottlenecks as memory grows, hindering real-time agent interactions. We propose SwiftMem, a query-aware agentic memory system that achieves sub-linear retrieval through specialized indexing over temporal and semantic dimensions. Our temporal index enables logarithmic-time range queries for time-sensitive retrieval, while the semantic DAG-Tag index maps queries to relevant topics through hierarchical tag structures. To address memory fragmentation during growth, we introduce an embedding-tag co-consolidation mechanism that reorganizes storage based on semantic clusters to improve cache locality. Experiments on LoCoMo and LongMemEval benchmarks demonstrate that SwiftMem achieves 47$times$ faster search compared to state-of-the-art baselines while maintaining competitive accuracy, enabling practical deployment of memory-augmented LLM agents.