🚀 MERA: Model Evolution and Routing with Skill Adaptation for Agentic Systems at Scale How can we make small models stronger for on-device agent deployment? MERA uses stronger models to guide an iterative loop of RL/GRPO, skill learning, and router optimization. Student failures become verified demonstrations, reusable SkillBook procedures, and LoRA updates, helping the small model take on more work over time. 🔥 Results: Qwen2.5-Coder-1.5B: 28.7% → 49.7% coding pass Qwen3.5-2B on TAU-2: 14/35 → 18/35 Fine-tuned 2B matches an unadapted 4B model Don’t just route around small models. Evolve them. 📄 arxiv.org/abs/2608.10333 💻 github.com/yh-yao/MERA-Evolv…

Kimi K3, the latest frontier by Moonshot AI is now available on @commonstack_ai! It is packed with 2.8T parameters with KDA enabling up to 6.3x faster in decoding. Built for self evolving and long horizon agentic tasks. Give the moon a shot 🌕

Kimi K3, the latest frontier by Moonshot AI is now available on @commonstack_ai! It is packed with 2.8T parameters with KDA enabling up to 6.3x faster in decoding. Built for self evolving and long horizon agentic tasks. Give the moon a shot 🌕

Excited to introduce AgenticTrading! 🚀 An open-source experimental playground for LLM-powered trading agents. Build, test, and deploy agents that reason, trade, and perform in realistic market environments. We are thrilled to collaborate with Dr. Xiaoyang Liu's group at Columbia University to push the boundaries of AI in finance. Stay tuned for our findings. Explore our platform: agentic-trading-lab.vercel.a… Read our Medium post: medium.com/@kuailedefcl/agen… GitHub Repo: github.com/Open-Finance-Lab/… We welcome more collaborators to join us! Let's shape the future of open finance together. 🌟 #AgenticTrading #AIinFinance #LLM #AlgorithmicTrading #OpenSource

Excited to share that TwinRouterBench has been accepted to the #RLEval Workshop at #CAIS2026 🎉 As LLM apps become long-horizon agents, one request can trigger many model calls across planning, tool use, retrieval, coding, and verification. That makes per-step LLM routing a core infrastructure problem: sending each call to the cheapest sufficient model without breaking downstream success. TwinRouterBench introduces: ⚡ Static track: 970 router-visible prefixes from 520 instances across SWE-bench, BFCL, mtRAG, QMSum, and PinchBench 🚀 Dynamic track: live SWE-bench Verified evaluation with official task resolution + realized API spend Key result: a router trained on static labels achieves comparable SWE-bench resolve rate while cutting API cost by ~53% vs. an unrouted Opus 4.6 baseline. Paper: arxiv.org/html/2605.18859v1 Code: github.com/CommonstackAI/Twi… Dataset: huggingface.co/datasets/Amor… Website: commonstackai.github.io/Twin… #LLM #AgenticAI #LLMRouting #Benchmark #SWEBench

Excited to share that TwinRouterBench has been accepted to the #RLEval Workshop at #CAIS2026 🎉 As LLM apps become long-horizon agents, one request can trigger many model calls across planning, tool use, retrieval, coding, and verification. That makes per-step LLM routing a core infrastructure problem: sending each call to the cheapest sufficient model without breaking downstream success. TwinRouterBench introduces: ⚡ Static track: 970 router-visible prefixes from 520 instances across SWE-bench, BFCL, mtRAG, QMSum, and PinchBench 🚀 Dynamic track: live SWE-bench Verified evaluation with official task resolution + realized API spend Key result: a router trained on static labels achieves comparable SWE-bench resolve rate while cutting API cost by ~53% vs. an unrouted Opus 4.6 baseline. Paper: arxiv.org/html/2605.18859v1 Code: github.com/CommonstackAI/Twi… Dataset: huggingface.co/datasets/Amor… Website: commonstackai.github.io/Twin… #LLM #AgenticAI #LLMRouting #Benchmark #SWEBench