Happy to see that our chemical language foundation model, MoLFormer is highlighted in Nature Computational Science. In addition to showing competitive performance in standard prediction benchmarks, it also shows first-of-a-kind emergent behavior with scaling, e.g. learning of geometry and taste

Good article on LLMs at Forbes. The media are starting to agree with my much-criticized statements about LLMs. "LLMs as they exist today will never replace Google Search. Why not? In short, because today’s LLMs make stuff up." forbes.com/sites/robtoews…

If you're using GPT-3 or any other LLMs read this: 1. Don't want it to hallucinate? 2. Need attribution for generated answers? 3. Have access to proprietary data that you want to index yourself and generate answers from it? Use PrimeQA! We added "retrieve" and "read" mode.🧵

We can all agree we’re at a unique and evolutionary moment in AI, with enterprises increasingly turning to this technology’s transformative power to unlock new levels of innovation and productivity. At #Think2023, IBM unveiled watsonx. Learn more: newsroom.ibm.com/2023-05-09-IBM…

We are releasing `v0.5.4` version of the transition-amr-parser. Now with document-level AMR parsing, instalable from PyPI, shipped with trained checkpoints and SoTA performance. github.com/IBM/transition…

Self-RAG in LlamaIndex 🦙 We’re excited to feature Self-RAG, a special RAG technique where an LLM can do self-reflection for dynamic retrieval, critique, and generation (Akari Asai et al.). It’s implemented in LlamaIndex 🦙 as a custom query engine with

IBM presents API-BLEND A Comprehensive Corpora for Training and Benchmarking API LLMs There is a growing need for Large Language Models (LLMs) to effectively use tools and external Application Programming Interfaces (APIs) to plan and complete tasks. As such, there is

AAAI nuclear-workshop.github.io workshop Neuro-Symbolic Learning and Reasoning in the Era of Large Language Models Gary Marcus talk on “No AGI without Neurosymbolic AI.” Asim Munawar Artur d'Avila Garcez Francesca Rossi

Are you building and evaluating RAG systems? Presenting InspectorRAGet arxiv.org/abs/2404.17347 a platform for easily analyzing overall performance, instance level analysis, comprehensive metrics, and multiple models and more!

Granite 3.0 is our latest update for the IBM foundation models. The 8B and 2B models outperform strong competitors with similar sizes. The 1B and 3B MoE use only 400M and 800M active parameters to target the on-device use cases. Our technical report provides all the details you

🎉Today, we're pleased to announce the release of the Granite 3.0 model family, the latest open-licensed, general purpose LLMs from IBM 🎉 These have been a labor of love for my team at IBM Research, working closely with a host of collaborators across the company. We're excited

Announcing "IBM SWE-Agent 1.0", from my team IBM Research , the first SWE-Agent built only on top of open-source models while achieving competitive performance (23.7%) compared to frontier LLM-agents on SWE-Bench. More details in this blog: ibm.biz/ibm_swe

We released best-in-class Apache 2.0 licensed models for detecting general harm and RAG hallucinations as part of the Granite 3.0 release! Read more: linkedin.com/pulse/ibm-open… Documentation: ibm.com/granite/docs/m… Hugging Face: huggingface.co/collections/ib… Try them out!

🚨 New Dataset Alert🚨 We introduce ACP Bench. A question-answering style dataset that evaluates AI-model's ability to reason about Action, Change, and Planning. Checkout 🔗 ibm.github.io/ACPBench/ 📄 arxiv.org/abs/2410.05669

Look at those beautiful Granite Guardian safety vests! #brand #bootleg The Granite Guardian technical report is now on arXiv: arxiv.org/abs/2412.07724 Give it a read to see how the model is state-of-the-art in detecting harmful or hallucinated prompts and responses.

We, at IBM, released a new dataset for evaluating nested API sequencing and its on huggingface now.

Putting It All into Context: Simplifying Agents with LCLMs Putting all the core code in the context often leads to better performance on SWE-bench than using agent scaffolding

🎉 Excited to share our ACL 2025 paper: 🤖R2D2: Remembering, Replaying and Dynamic Decision Making with a Reflective Agentic Memory 🧠 📄 Paper: arxiv.org/abs/2501.12485 📍Poster: Hall 4/5, Session 4 Wednesday, July 30 11:00-12:30 🧵👇

New IBM Research paper builds a judge for tool calls that makes tool using LLMs more accurate. It reports up to 25% higher accuracy on tool calling. Current judges rate normal text, not tool calls, so they miss wrong names, bad or missing parameters, and extra calls. The

An exciting milestone for AI in science: Our C2S-Scale 27B foundation model, built with Yale University and based on Gemma, generated a novel hypothesis about cancer cellular behavior, which scientists experimentally validated in living cells.  With more preclinical and clinical tests,