WizardLM (@wizardlm_ai) 's Twitter Profile
WizardLM

@wizardlm_ai

WizardLM, WizardCoder, WizardMath.

Evol-Instruct, Arena Learning, RLEIF.

ID: 1450726326818639872

calendar_today20-10-2021 07:31:43

347 Tweet

12,12K Followers

677 Following

Alpay Ariyak (@alpayariyak) 's Twitter Profile Photo

If you’ve spoken to me since the official announcement of WizardLM-2 in April, there’s a 99% chance I was in your ear rambling about how much I was looking forward to the paper detailing their new training data synthesis pipeline - today is my Christmas Thank you WizardLM

Yam Peleg (@yampeleg) 's Twitter Profile Photo

Don't sleep! WizardLM just dropped the (probably) best data generation method ever known to mankind! I didn't get to read it all out yet (later on today), but from what Iv'e seen the main ideas are: 1. Ask a model to "make the task harder" (WizardLM-1) 2. BUT also ask a model

Don't sleep!

WizardLM just dropped the (probably) best data generation method ever known to mankind!

I didn't get to read it all out yet (later on today), but from what Iv'e seen the main ideas are:

1. Ask a model to "make the task harder" (WizardLM-1)
2. BUT also ask a model
fullstack (@davidfswd) 's Twitter Profile Photo

Exciting! Evol Instruct v2 paper is out! Microsoft Research WizardLM just released Auto-Evol-Instruct! (I consider this type of work pre/proto-agi, but my views are sometimes radical) arxiv.org/pdf/2406.00770

Exciting! Evol Instruct v2 paper is out! 

Microsoft Research WizardLM just released Auto-Evol-Instruct!

(I consider this type of work pre/proto-agi, but my views are sometimes radical)

arxiv.org/pdf/2406.00770
Marco Mascorro (@mascobot) 's Twitter Profile Photo

The WizardLM team is back, with Evol-Instruct V2 (actually more like "Auto" Evol-Instruct), which is one of core components for WizardLM-2 Thank you WizardLM team.

Aidan McLaughlin (@aidan_mclau) 's Twitter Profile Photo

-- <big_model_smell> benchmark -- Aidan Bench measures creativity, reliability, attention, and instruction following. >mistral large 2 wins by a lot??? >gpt-4o sucks confirmed >sonnet-3.5 remains very strong >gpt-4-0314 shows old man strength github.com/aidanmclaughli…

-- &lt;big_model_smell&gt; benchmark --

Aidan Bench measures creativity, reliability, attention, and instruction following.

&gt;mistral large 2 wins by a lot???
&gt;gpt-4o sucks confirmed
&gt;sonnet-3.5 remains very strong
&gt;gpt-4-0314 shows old man strength

github.com/aidanmclaughli…
Qingfeng Sun (@victorsungo) 's Twitter Profile Photo

Congrats this impressive contribution to OSS community! Also excited to see the state-of-the-art Hermes-3 models also leverage our Evol-Instruct to empower their complex instruction following capacities.

Congrats this impressive contribution to OSS community! 

Also excited to see the state-of-the-art Hermes-3 models also leverage our Evol-Instruct to empower their complex instruction following capacities.
Microsoft Research (@msftresearch) 's Twitter Profile Photo

Learn what’s next for AI at Research Forum on Sept. 3; WizardArena simulates human-annotated chatbot games; MInference speeds pre-filling for long-context LLMs via dynamic sparse attention; Reef: Fast succinct non-interactive zero-knowledge regex proofs. msft.it/6019l4Qv9

Learn what’s next for AI at Research Forum on Sept. 3;  WizardArena simulates human-annotated chatbot games; MInference speeds pre-filling for long-context LLMs via dynamic sparse attention; Reef: Fast succinct non-interactive zero-knowledge regex proofs. msft.it/6019l4Qv9
Ting-En Lin (@tnlin_tw) 's Twitter Profile Photo

🚀 Excited to introduce MMEvol, which improves MLLM through complex and diverse instruction with perceptual, cognitive, and interactive evolution. 🌟 Achieving a 3.1% accuracy boost across 13 VL tasks. Code and data will be released soon; stay tuned! 📄 huggingface.co/papers/2409.05…

🚀 Excited to introduce MMEvol, which improves MLLM through complex and diverse instruction with perceptual, cognitive, and interactive evolution.
🌟 Achieving a 3.1% accuracy boost across 13 VL tasks.
Code and data will be released soon; stay tuned!
📄 huggingface.co/papers/2409.05…
Lucas Atkins (@lucasatkins7) 's Twitter Profile Photo

We are open sourcing our EvolKit pipeline that was instrumental in the creation of supernova, under MIT license. This was heavily inspired by the AutoEvol paper from WizardLM, and is a tremendously powerful tool for creating complex datasets. Find it here:

qnguyen3 (@stablequan) 's Twitter Profile Photo

Impressed by Arcee.ai team's work. Proud to open-source EvolKit: framework for evolving instruction data with OPEN-SOURCE models. Inspired by WizardLM, result of my month-long effort. GitHub: github.com/arcee-ai/EvolK… Don't forget SuperNova too! 🥳

WizardLM (@wizardlm_ai) 's Twitter Profile Photo

Congrats! We are dedicated to innovating synthetic training techniques, drawing inspiration from theory of evolution. The previous Evol-Instruct focused on evolving higher-value instructions from the instruction side. This work, Arena Learning, emphasizes the evolution of

elvis (@omarsar0) 's Twitter Profile Photo

Agentic Information Retrieval This paper provides a good introduction to agentic information retrieval, which is shaped by the capabilities of LLM agents. I've been developing with this paradigm recently and it does offer lots of interesting ways to optimize retrieval systems.

Agentic Information Retrieval

This paper provides a good introduction to agentic information retrieval, which is shaped by the capabilities of LLM agents.

I've been developing with this paradigm recently and it does offer lots of interesting ways to optimize retrieval systems.
kaeru-0.5B (@mryo39) 's Twitter Profile Photo

GENIAC phase2にて、日本語のローカルLLMを使ってEvol-Instructによるデータセット構築に取り組んだ際の記事を公開しました。(3件目/全4件) zenn.dev/matsuolab/arti…

WizardLM (@wizardlm_ai) 's Twitter Profile Photo

🚀New approach from WaveCoder Team for optimizing code LLMs. The novel feature tree based framework, inspired by AST and Evol-Instruct to modeling semantic relationships, generates more diverse data. The EpiCoder hits SOTA in both challenge file and function benchmarks.

Mengkang Hu (@aaron_mkhu) 's Twitter Profile Photo

🎉 Thrilled to share our paper accepted by #KDD2025! 🌟AgentGen🌟: An automated environment and task generator that enhances LLM-based agents' planning abilities through diverse, difficulty-controlled synthetic trajectory data. 👇🏻agent-gen.github.io