OpenHands: Open Platform for AI Software Agents and What It Means for Finance Automation
OpenHands' CodeAct agent scores 26% on SWE-Bench Lite, showing what AI agents reliably do today. Finance automation should start tightly scoped, not autonomous.
#developers
Разработчици
Ресурси за разработчици, API и документация за интеграция на финансови инструменти
ShieldAgent: Verifiable Safety Policy Reasoning for LLM Agents
ShieldAgent hits 90.4% accuracy on agent attacks with 64.7% fewer API calls by using probabilistic rule circuits instead of LLM guardrails.
Fine-Tuning vs. RAG: Why Retrieval Wins for Injecting New Knowledge into LLMs
RAG hits 0.875 accuracy on post-cutoff facts while fine-tuning plateaus at 0.504. For agents needing frequent ledger updates, retrieval beats fine-tuning.
Gorilla: How Retrieval-Aware Training Reduces LLM API Hallucinations from 78% to 11%
Gorilla's Retriever-Aware Training cuts LLM API hallucination rates from 78% to 11%, making tool calls reliable enough for finance agents that write entries.
SWE-agent: How Interface Design Unlocks Automated Software Engineering
SWE-agent's Agent-Computer Interfaces lifted GPT-4 Turbo from raw shell to 12.47% on SWE-bench, a 10.7-point gain from interface design alone.
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
SWE-bench tests language models on 2,294 real GitHub issues; at publication Claude 2 resolved only 1.96%. Retrieval and patch-length limits shape coding agents.
Toolformer: Самоконтролирано използване на инструменти и неговите граници за финансовите ИИ
Toolformer учи 6.7B модел да вика API чрез филтриране по перплексия, побеждавайки GPT-3 175B по аритметика. Едноетапният му дизайн блокира верижни заявки към регистъра.