ETH Zurich
PulseAugur coverage of ETH Zurich — every cluster mentioning ETH Zurich across labs, papers, and developer communities, ranked by signal.
- 2026-07-26 product_launch ETH Zurich has developed a new sensor chip capable of detecting deepfakes by analyzing imperceptible digital imperfections. source
6 day(s) with sentiment data
-
AI's climate claims questioned amid Nvidia CEO probe · 2 sources tracked
A recent newsletter highlights concerns that the purported climate benefits of AI are overstated, with industries failing to adequately leverage the technology for emissions reduction. Separately, Nvidia is reportedly i…
-
NVIDIA's open-source tools accelerate 6G network research · 2 sources tracked
NVIDIA is advancing the development of next-generation 6G networks through its suite of open-source tools, including NVIDIA AI Aerial and NVIDIA Sionna. These tools, which have seen a tenfold increase in adoption over t…
-
PhD grad seeks advice on industry-to-academia transition for ML faculty roles
A PhD graduate with research experience in LLM efficiency and RL is seeking advice on transitioning from industry back to academia. Despite publishing in top ML and NLP conferences, the individual lacks experience in gr…
-
New Rowhammer exploits target AMD Zen chips and DDR5 SDRAM
Researchers have identified new Rowhammer exploits targeting AMD Zen chips and DDR5 SDRAM, with one exploit named Phoenix successfully bypassing existing mitigations. These findings, presented by researchers at ETH Züri…
-
ETH Zurich trains robot G1 to climb ladders with 10 minutes of data
Researchers at ETH Zurich have developed a new method for teaching robots complex physical tasks, such as climbing ladders and navigating challenging terrain. This approach utilizes contact prediction and requires only …
-
AI coding agents' effectiveness tied to context management, not just models
The effective use of AI coding agents like Claude Code, Codex, and Gemini CLI hinges on managing the context they process, which significantly impacts session quality and cost. Developers increasingly rely on these tool…
-
Ecologist studies natural rainforest regeneration via animal seed dispersal
Ecologist Daisy Dent is studying how tropical forests naturally regenerate, focusing on the crucial role of animals in dispersing seeds. Her research in Panama, spanning over a century of data, reveals that diverse ecos…
-
NLP research roundup: 4 key papers from September 2026
This article highlights four significant research papers from September 2026 in the Natural Language Processing (NLP) field. It aims to inform readers about the latest findings and developments within NLP. The featured …
-
Darwin-180B-RSI model tops legal benchmarks without domain training · 1 source tracked
VIDRAFT's Darwin-180B-RSI, a 180-billion-parameter large language model, has achieved top performance on Swiss legal reasoning benchmarks, including LEXam and LEXam-hard. Notably, the model accomplished this without any…
-
AI model tops Swiss law exam benchmark without legal training
An open-weight 180B model named Darwin-180B-RSI, developed by the Korean startup VIDRAFT, has achieved top rankings on legal reasoning benchmarks, despite never being trained on legal data. The model utilizes a novel ap…
-
Heliox podcast explores AI hallucinations and geometric solutions
A new episode of the Heliox podcast explores how AI models hallucinate due to their simplified understanding of the world, likening it to seeing in flat shadows. The episode discusses how incorporating curved geometry, …
-
EPFL researchers partner with Proton to integrate Apertus 1.5 into Lumo AI assistant
Researchers from EPFL, ETH Zurich, and the Swiss National Supercomputing Centre have partnered with Proton to integrate their Apertus 1.5 model into Proton's AI assistant, Lumo. This collaboration aims to provide resear…
-
ETH Zurich unifies discrete diffusion models with new proof
Researchers at ETH Zurich have published a preprint demonstrating that four leading discrete diffusion model methods can be unified under a single objective function. This new proof not only consolidates these distinct …
-
Snowden, Meyer, Subramanian, Brekke to discuss AI, surveillance, free speech
Edward Snowden, Marguerite Meyer, Vikram Subramanian, and Jaya Brekke are participating in a discussion about AI, surveillance, and freedom of speech. The event is scheduled for September 24th at ETH Zurich.
-
TERMy: New terminal assistant bypasses LLMs for local command execution
A new terminal assistant called TERMy has been developed, aiming to provide efficient command-line operations without relying on large language models. The project, initiated due to rising AI costs and privacy concerns,…
-
Claude Certified Architect exam takers share preparation tips and experiences
Two individuals share their experiences preparing for and passing the Claude Certified Architect - Foundations exam. One author details their preparation process, emphasizing practice tests and understanding the exam's …
-
ETH Zurich pushes LLMs to 1-bit quantization for small devices
Researchers from ETH Zurich have developed innovative techniques to drastically reduce the size of large language models (LLMs) for deployment on resource-constrained devices. Their work, presented at IJCAI 2026, focuse…
-
Pakistan's JudgeGPT trial boosts case resolution by 6.3% with AI
A large-scale experiment in Pakistan involving a custom AI tool called JudgeGPT has shown a 6.3% increase in resolved cases among trial judges. The tool, developed in consultation with the judiciary, combines OpenAI's G…
-
Apertus-15 champions open-source AI infrastructure with transparent training
Apertus-15, a project from ETH Zurich's AI Center, is developing open-source AI infrastructure focused on transparency and trustworthiness. Unlike frontier models from private companies, Apertus aims to provide a founda…
-
Bounty offered for LLM-powered deanonymization pipelines
A $500 bounty has been offered for the development of large-scale attack pipelines capable of deanonymizing individuals using LLMs. This initiative is inspired by research from Anthropic and ETH Zurich, which demonstrat…