Research priority queue

研究优先级队列

这是一台研究排队器：哪些先深读，哪些先补证据，哪些只做背景观察；最终判断必须回到原始证据、报告和 benchmark。

30展示项目

50%时间权重

239分析项目

Queue	Project	Priority	Time	Mechanism	Evidence	Category	Created	Activity
1	AgentEvolver modelscope/AgentEvolver	83	76	100	100	Agent 进化框架	2025-11-13	2026-04-01
2	DARWIN ZJU-LLM-Safety/DARWIN	77	94	75	62	安全策略进化	2026-04-18	2026-05-07
3	LLM-Self-Judge OPPO-Mente-Lab/LLM-Self-Judge	76	87	85	62	自评判训练	2026-03-12	2026-03-24
4	openevolve algorithmicsuperintelligence/openevolve	71	57	79	100	进化式代码优化	2025-05-15	2026-03-18
5	SCOPE JarvisPei/SCOPE	68	79	67	62	上下文/Prompt 进化	2025-12-18	2026-03-26
6	science-codeevolve inter-co/science-codeevolve	66	74	59	68	科学代码进化	2025-10-14	2026-04-08
7	SE-Agent JARVIS-Xs/SE-Agent	64	48	77	100	代码智能体自进化	2025-07-11	2025-09-23
8	EvoMap Evolver EvoMap/evolver created_at unavailable; time score capped and based on activity/observation only	62	45	93	62	Self-Evolving Memory and Reasoning Map Framework	unknown	2026-05-31
9	Agent Lightning microsoft/agent-lightning created_at unavailable; time score capped and based on activity/observation only	61	45	95	56	Reinforcement-Learning Agent Training Framework	unknown	2026-05-29
10	Hermes Agent Self-Evolution NousResearch/hermes-agent-self-evolution created_at unavailable; time score capped and based on activity/observation only	61	45	100	62	On-Policy RL Self-Evolution Pipeline for Agent Models	unknown	2026-05-31
11	holaOS holaboss-ai/holaOS created_at unavailable; time score capped and based on activity/observation only	60	45	95	56	Long-Horizon Agent Environment	unknown	2026-05-27
12	Self Evolve OpenClaw Playground longmans/self-evolve created_at unavailable; time score capped and based on activity/observation only	58	45	97	56	Self-Evolving OpenClaw Workflow Playground and Benchmark Harness	unknown	2026-06-01
13	Meta-Harness (Stanford IRIS) stanford-iris-lab/meta-harness created_at unavailable; time score capped and based on activity/observation only	58	45	87	56	Meta-Harness Framework and Reference Experiments	unknown	2026-05-26
14	AegisLLM zikuicai/aegisllm created_at unavailable; time score capped and based on activity/observation only	57	45	97	56	Self-Reflective Multi-Agent Defense System	unknown	2026-05-30
15	REINS Self-Improving Model Framework pegasi-ai/reins created_at unavailable; time score capped and based on activity/observation only	57	45	95	56	Self-Improving Agent Policy Framework and Training Harness	unknown	2026-06-01
16	EvoAgentX EvoAgentX/EvoAgentX created_at unavailable; time score capped and based on activity/observation only	57	29	97	86	自进化 Agent 生态系统	unknown	2026-01-01
17	Darwin Mobile Agent ai-agents-2030/darwin-mobile-agent created_at unavailable; time score capped and based on activity/observation only	57	45	100	56	Mobile Agent Self-Evolution Framework	unknown	2026-05-29
18	NexAgent gofenix/nex-agent created_at unavailable; time score capped and based on activity/observation only	57	45	95	56	Elixir/OTP Self-Evolving Agent Runtime	unknown	2026-05-26
19	SEAD Da1yuqin/SEAD created_at unavailable; time score capped and based on activity/observation only	57	45	97	56	Self-Evolving Agent Design Benchmark	unknown	2026-05-29
20	InfiAgent InfiAgent/InfiAgent created_at unavailable; time score capped and based on activity/observation only	57	45	87	56	Framework for Self-Improving Agent Loops	unknown	2026-05-29
21	Skills Vote Evolution Benchmark MemTensor/skills-vote created_at unavailable; time score capped and based on activity/observation only	57	45	85	56	Self-Evolving Skill Selection and Benchmark Pipeline	unknown	2026-05-31
22	AutoHarness aiming-lab/AutoHarness created_at unavailable; time score capped and based on activity/observation only	56	45	87	56	Automated Agent Harness Engineering Framework	unknown	2026-05-30
23	Self-Improving Agent BerriAI/self-improving-agent created_at unavailable; time score capped and based on activity/observation only	56	45	97	56	Self-Improving Coding Agent Loop	unknown	2026-05-29
24	SICA Self-Improving Coding Agent MaximeRobeyns/self_improving_coding_agent created_at unavailable; time score capped and based on activity/observation only	55	45	87	56	Self-Improving Coding Agent	unknown	2026-05-26
25	SkillOpt microsoft/SkillOpt created_at unavailable; time score capped and based on activity/observation only	55	45	75	56	Self-Evolving Agent Skill Optimizer	unknown	2026-05-28
26	Yunjue Agent YunjueTech/Yunjue-Agent created_at unavailable; time score capped and based on activity/observation only	54	35	97	56	In-Situ Self-Evolving Agent System	unknown	2026-02-11
27	self-evolving-agent RangeKing/self-evolving-agent created_at unavailable; time score capped and based on activity/observation only	54	45	85	56	OpenClaw Self-Evolving Skill	unknown	2026-05-26
28	DSPy stanfordnlp/dspy	53	29	59	100	声明式 Prompt 优化	2023-01-09	2026-05-23
29	OpenAI Swarm openai/swarm created_at unavailable; time score capped and based on activity/observation only	53	45	57	56	Experimental Multi-Agent Orchestration Framework	unknown	2026-05-29
30	HyperSpell OpenClaw Memory Engine hyperspell/hyperspell-openclaw created_at unavailable; time score capped and based on activity/observation only	52	45	67	56	OpenClaw Memory and Context Enhancement Runtime	unknown	2026-06-01

Method

这个队列怎么用

先用它决定下一步行动：优先深读、重点跟踪、补证据或背景观察。数字只是研究排队器的粗粒度线索，不是最终学术结论。

怎么用这个队列

优先深读 1

80+ 分机制、证据和近期性都强，适合优先阅读源码和项目报告。

AgentEvolver

重点跟踪 3

70-79 分方向清楚，但还需要补足报告、对照实验或长期活跃度。

DARWIN · LLM-Self-Judge · openevolve

补证据 7

60-69 分可能有价值，当前更适合回到原始仓库和报告验证机制。

SCOPE · science-codeevolve · SE-Agent

背景观察 19

<60 分作为生态背景，不应抢占当前研究优先级。

Self Evolve OpenClaw Playground · Meta-Harness (Stanford IRIS) · AegisLLM

当前队列的机制信号覆盖

自进化闭环

25 60 avg · AgentEvolver

安全治理

5 62 avg · DARWIN

查看评分权重和证据边界

研究优先级队列不是历史影响力榜。时间新近性占 50%，自进化机制强度占 20%，证据链完整度占 15%，社区动量占 10%，实用适配度占 5%。GitHub API 无法确认创建时间的项目会被标记 caveat，时间分最高封顶 45/100，避免本地 mirror 或采集时间制造“假新项目”。

Dimension	Weight	Meaning
Time	50%	创建时间新近性 + 最近活动；创建时间缺失时封顶。
Mechanism	20%	是否真的包含进化、评估器、记忆/技能、训练/数据循环。
Evidence	15%	是否有 raw、classification、公开报告、GitHub API 或本地镜像证据。
Adoption	10%	log-scaled stars，只作为动量信号，不主导复核顺序。
Usefulness	5%	是否适合工程使用、评测、harness、memory/skill 或自进化实践。