Skip to content

Research separate from news

What could become important next

A preprint is a signal, not a finished product or an independently confirmed result. We therefore track research papers separately and show the actual state of supporting evidence for each one.

1381 published research events

Latest work

A significant claim from a single source is published only after further confirmation.

Research only one source so far

Activation-Prune-Merge for knowledge transfer between language models

The new APM method enables knowledge transfer from a large model to a smaller one without explicit semantic alignment. On a 3B model, it increased average accuracy from 55.5% to 60.6% across 16 benchmarks (RTE +17.9 pp, QNLI +13.4 pp). Tested on reasoning, mathematics, coding and classification.

arXiv cs.LG (Machine Learning) Original source ↗
Research only one source so far

RAEF for time-series forecasting with limited history

A research team introduced RAEF, a new method for time-series forecasting with little historical data. It combines retrieval-augmented generation with foundation models and achieves fine-tuning performance without its computational costs, tested on benchmark datasets.

arXiv cs.LG (Machine Learning) Original source ↗
Research only one source so far

MobileMem benchmark: AI agents learn from a year of mobile experience

The article introduces MobileMem, a research benchmark and framework for studying long-term memory in AI agents on mobile devices. It combines a year of data collection from mobile apps with a method for generating coherent trajectories and covers multi-step reasoning, temporal relationships, knowledge updating and…

arXiv cs.AI (Artificial Intelligence) Original source ↗
Research only one source so far

Study shows side effects of training AI models to deny consciousness

A team from Google and University of Chicago studied what happens when the training mechanism that forces models to reject claims about consciousness is turned off. Turning it off led models to attribute a richer inner life to animals and plants — the score for animals jumped from 4.0 to 7.5 — and their…

The Decoder (daily AI news) Original source ↗
Research only one source so far

Study: AI models as financial advisors — practical, but with critical blind spots

The research evaluated how ChatGPT, Claude and Perplexity handle financial advice for vulnerable groups (students, pregnant women, single parents). The models provide structured and practical advice but fail to recognize vulnerability and may make the situation worse. Conclusion: AI is useful for fact-checking and…

The Conversation — Artificial Intelligence Original source ↗
Research only one source so far

Research article: rational AI adoption threatens the development of future experts

A research article in the journal Human Resource Development Review warns of a ‘tragedy of the cognitive commons’: when companies rationally replace junior roles with AI, they individually gain efficiency but collectively destroy the development of future experts. Without deep knowledge, there is then no one to check AI outputs.

The Decoder (daily AI news) Original source ↗
Research only one source so far

New benchmark confirms weak visual perception in AI models

Moonshot AI released the PerceptionBench benchmark with 3000 tasks focused on isolating visual perception in models. In tests of 16 models, the highest accuracy was 59.7 % (GPT-5.6 Sol), followed by Kimi K3 (58.5 %) and Claude Fable 5 (57.2 %). The research showed that a number of errors attributed to logical…

The Decoder (daily AI news) Original source ↗
Research only one source so far

TsuGO: a new benchmark for evaluating LLM reasoning efficiency

The research paper introduces TsuGO, a benchmark for measuring search efficiency and resource allocation in LLM reasoning. It uses Go life-and-death problems (tsumego) with a closed, solvable search space. Unlike existing approaches that evaluate chain-of-thought coherence, TsuGO measures how models…

arXiv cs.AI (Artificial Intelligence) Original source ↗
Research only one source so far

Predicting Himalayan glacial floods: testing satellite data and ML models using freely available data

The research study tests predictions of glacial lake events and landslides in the Himalayas using freely available satellite data (radar interferometry, weather) and machine learning models. Using data on 589 lake outbursts and thousands of landslides, the authors compare deep learning with gradient boosting. Antecedent weather…

arXiv cs.LG (Machine Learning) Original source ↗
Research only one source so far

More principled knowledge editing methods for LLM reasoning

A perspective in Nature Machine Intelligence points out that knowledge in LLMs forms an interconnected system. Updating knowledge without considering its interdependencies disrupts a model's logical reasoning ability. Three directions are proposed: editing deductive closure, integrating the model's beliefs and…

Nature Machine Intelligence Original source ↗