AgentForge is a multi-agent LLM system that mandates sandboxed execution verification for code changes and reports 40% resolution on SWE-Bench Lite.
Evaluating large language models trained on code,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
AgentForge: Execution-Grounded Multi-Agent LLM Framework for Autonomous Software Engineering
AgentForge is a multi-agent LLM system that mandates sandboxed execution verification for code changes and reports 40% resolution on SWE-Bench Lite.