terminalbench

TerminalBench-2 benchmark for agentic coding and terminal-based task execution

Name	Categories	Topics	References	Credibility	Title	NavigationOrder	Eleventy	TemplateEngineOverride	RelatedNotes
20260423_meta_harness_end_to_end_optimization	paper_analysis	context_engineering, agent_architecture, llm, reasoning, rag	stanford_iris_lab, terminalbench_2, krafton	100	Meta-Harness: End-to-End Optimization of Model Harnesses	[object Object]	[object Object]	md
20260609_socratic_swe_self_evolving_coding_agents_trace_derived_skills	paper_analysis	self_evolving_agents, memory_mechanism, skill_curation, reinforce_learning, reasoning	sjtu, swebench_verified, terminalbench_2, grpo, skillos	100	Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills	[object Object]	[object Object]	md	skillos_learning_skill_curation_self_evolving_agents, skillopt_executive_strategy_self_evolving_agent_skills, aflow_automating_agentic_workflow_generation, scaling_harness_agentic_ai, meme_multi_entity_evolving_memory_evaluation
20260624_evoarena_tracking_memory_evolution	paper_analysis	memory_mechanism, agent_architecture, llm, evaluation, embodied_ai	nus, ntu, terminalbench_2	100	EvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments	[object Object]	[object Object]	md	Self-Harness: Harnesses That Improve Themselves
20260624_self_harness	paper_analysis	agent_architecture, self_evolving_agents, llm, reasoning, test_time_scaling	shanghai_ai_lab, terminalbench_2	100	Self-Harness: Harnesses That Improve Themselves	[object Object]	[object Object]	md