Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal human cancers and shows resistance to any therapeutic strategy used. Here we tested small-molecule inhibitors targeting chromatin regulators as possible therapeutic agents in PDAC. We show that JQ1, an inhibitor of the bromodomain and extraterminal (BET) family of proteins, suppresses PDAC development in mice by inhibiting both MYC activity and inflammatory signals. The histone deacetylase (HDAC) inhibitor SAHA synergizes with JQ1 to augment cell death and more potently suppress advanced PDAC. Finally, using a CRISPR-Cas9-based method for gene editing directly in the mouse adult pancreas, we show that de-repression of p57 (also known as KIP2 or CDKN1C) upon combined BET and HDAC inhibition is required for the induction of combination therapy-induced cell death in PDAC. SAHA is approved for human use, and molecules similar to JQ1 are being tested in clinical trials. Thus, these studies identify a promising epigenetic-based therapeutic strategy that may be rapidly implemented in fatal human tumors.
Wie kann man Schüler optimal fördern? Welche Lernmethoden und Unterrichtsstrategien funktionieren wirklich? Auf Basis pädagogisch-psychologischer Erkenntnisse liefert dieser Band praxistaugliche Antworten. Er stellt Ergebnisse der Unterrichtsforschung vor, diskutiert Qualitätsmerkmale und gibt Tipps zur Gestaltung des eigenen Unterrichts. Fallbeispiele und Übungen helfen Studenten, Referendaren und Lehrern bei der Umsetzung in die Praxis.
Modern Recurrent Neural Networks have become a competitive architecture for 3D reconstruction due to their linear-time complexity. However, their performance degrades significantly when applied beyond the training context length, revealing limited length generalization. In this work, we revisit the 3D reconstruction foundation models from a Test-Time Training perspective, framing their designs as an online learning problem. Building on this perspective, we leverage the alignment confidence between the memory state and incoming observations to derive a closed-form learning rate for memory updates, to balance between retaining historical information and adapting to new observations. This training-free intervention, termed TTT3R, substantially improves length generalization, achieving a 2 $\times$ improvement in global pose estimation over baselines, while operating at 20 FPS with just 6 GB of GPU memory to process thousands of images. Code is available in https://rover-xingyu.github.io/TTT3R.
Agriculture was central to the historical development of the Levant, with food storage ensuring year-round availability and managing seasonal, environmental, and climatic challenges. Additionally, studying storage practices reveals insights into production organization, surplus management, and socio-economic structures linked to urbanization and trade. While archaeological and archaeobotanical research has advanced our understanding of Levantine agriculture, a comprehensive archaeobotanical synthesis of Bronze and Iron Age storage remains lacking. This study examines how storage contexts can be identified archaeologically and archaeobotanically, what was stored, how practices varied regionally and chronologically, and which additional practices play a role. New material from underrepresented regions, such as Kamid el-Loz in the central Levant (Lebanon), expands the ADEMNES dataset. Data from 18 sites with identified storage contexts and 43 sites with non-storage contexts are synthesized to provide a region-wide analysis. Based on this study, key criteria for identifying storage contexts include the presence of storage facilities or containers if preserved, substantial in situ concentrations of charred plant remains, predominance of easily preserved crops, largely cleaned assemblages, and potential evidence of storage pests. Multivariate analysis, namely Principal Component Analysis (PCA), shows that storage context is the main factor shaping plant assemblages, followed by chronological, regional, environmental, and socio-economic factors. In conclusion, this study offers a more systematic understanding of Bronze and Iron Age Levantine food storage practices from an archaeobotanical point of view.
Continual learning (CL) in large language models (LLMs) is an evolving domain that focuses on developing efficient and sustainable training strategies to adapt models to emerging knowledge and achieve robustness in dynamic environments. Our primary emphasis is on continual domain-adaptive pretraining, a process designed to equip LLMs with the ability to integrate new information from various domains while retaining previously learned knowledge. Since existing works concentrate mostly on continual fine-tuning for a limited selection of downstream tasks or training domains, we introduce a new benchmark designed to measure the adaptability of LLMs to changing pretraining data landscapes. We further examine the impact of model size on learning efficacy and forgetting, as well as how the progression and similarity of emerging domains affect the knowledge transfer within these models. Our findings uncover several key insights: (i) continual pretraining consistently improves <1.5B models studied in this work and is also superior to domain adaptation, (ii) larger models always achieve better perplexity than smaller ones when continually pretrained on the same corpus, (iii) smaller models are particularly sensitive to continual pretraining, showing the most significant rates of both learning and forgetting, (iv) continual pretraining boosts downstream task performance of GPT-2 family, (v) continual pretraining enables LLMs to specialize better when the sequence of domains shows semantic similarity while randomizing training domains leads to better transfer and final performance otherwise. We posit that our research establishes a new benchmark for CL in LLMs, providing a more realistic evaluation of knowledge retention and transfer across diverse domains.
