Current harnesses have a very rigid setup in how the LLM can manage its own context. We show that you can let LLMs edit the context direc...
Current harnesses have a very rigid setup in how the LLM can manage its own context. We show that you can let LLMs edit the context directly - enabling better long-horizon memory, FLOP efficient adaption, & more.
Don't fear compaction anymore!
More AI News
- 19:54 X: Rohan Paul (@rohanpaul_ai)
New Nvidia paper: AI models get sloppier as jobs get longer, even inside their context window, so number every item and split big jobs in...
New Nvidia paper: AI models get sloppier as jobs get longer, even inside their context window, so number every item and split big jobs into small chunks.Model size didn't guarantee reliability on long, repetitive jobsPicture an agent updating a huge invoice file line by line. It can read the whole file and still skip a line or update the wrong record.NVIDIA tested 7 open models on simple, repetitive jobs like adding numbers and sorting lists. Average accuracy was 62.8% lower on 128K-token jobs t
- 19:18 X: Testing Catalog (@testingcatalog)
Volantis has raised $88 million in Series A funding co-led by Lachy Groom and Abstract, with John Doerr, Susa Ventures, VXI Capital, and ...
Volantis has raised $88 million in Series A funding co-led by Lachy Groom and Abstract, with John Doerr, Susa Ventures, VXI Capital, and Triatomic Capital also participating.Volantis is building optical links inside chip packages, providing AI accelerators access to large pools of faster, lower-cost memory.The company reports that memory bandwidth and capacity per chip can increase by several times compared to current designs.Volantis aims for up to 10,000 tokens per second per user on models wi
- 19:15 The Decoder: AI News (RSS)
AI beats licensed accountants on speed and accuracy, but still can't close the books without supervision
AI models are faster, more accurate, and far cheaper than accountants at structured bookkeeping tasks. That's the finding of a study by Mercor in which 12 licensed CPAs with an average of five and a half years of experience worked through simplified tasks from the APEX Accounting Benchmark. Eighteen months ago, the best models still scored below the accountants' average of about 37 percent. Today, they solve the same tasks almost flawlessly.AI model accuracy on the study's accounting tasks compa
- 19:00 Databricks: Blog (RSS)
Real-Time Retail Intelligence: Building E-Commerce Recommendations with Lakebase and AI Search on Databricks
The opportunity: Personalization as a revenue engineEvery second a shopper spends on a fashion e-commerce app generates a stream of intent signals — searches, product views, wishlist additions, cart interactions. The platforms that convert those signals into relevant product recommendations in real time are the ones that win. Industry benchmarks show that effective personalization can lift conversion rates by 10–30% and increase average order value significantly.Yet building a production-grade r