eidolons/research/references.bib at main · Rynaro/eidolons · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
% Eidolons — Research BibTeX
% ═══════════════════════════════════════════════════════════════════════════
% Citations used across the Eidolons team. Each entry is referenced by
% individual Eidolon DESIGN-RATIONALE.md files and by the nexus
% research library index.
%
% Convention: key = FirstAuthorYearShortTitle (e.g. Shinn2023Reflexion)
% ═══════════════════════════════════════════════════════════════════════════

% ─── Self-correction, verification, bounded loops ──────────────────────────

@inproceedings{Shinn2023Reflexion,
  title     = {Reflexion: Language Agents with Verbal Reinforcement Learning},
  author    = {Shinn, Noah and Cassano, Federico and Berman, Edward and
               Gopinath, Ashwin and Narasimhan, Karthik and Yao, Shunyu},
  booktitle = {NeurIPS},
  year      = {2023},
  note      = {Foundational work for reflective agents. Eidolons adopt the
               pattern but enforce a bounded revision budget — CorrectBench 2025
               showed unbounded variants degrade output on open-ended tasks.}
}

@misc{CorrectBench2025,
  title   = {CorrectBench: Evaluating Self-Correction in Language Model Agents},
  author  = {{CorrectBench Authors}},
  year    = {2025},
  note    = {Empirical evidence for bounding reflection loops. Drives
             APIVR-$\Delta$'s 3-attempt cap and IDG's 1-pass gate.}
}

% ─── Documentation & synthesis ─────────────────────────────────────────────

@article{DocAgent2024,
  title   = {DocAgent: Multi-Role Documentation Generation for Codebases},
  author  = {{Meta FAIR}},
  journal = {arXiv},
  eprint  = {2504.08725},
  year    = {2024},
  note    = {CHT (Completeness/Helpfulness/Truthfulness) verification
             framework adopted by IDG's Gate phase. Also source of
             topological-ordering evidence for drafting sequence.}
}

% ─── Software engineering agents ───────────────────────────────────────────

@inproceedings{Yang2024SWEAgent,
  title     = {SWE-Agent: Agent-Computer Interfaces Enable Automated Software Engineering},
  author    = {Yang, John and Jimenez, Carlos and Wettig, Alexander and
               Lieret, Kilian and Yao, Shunyu and Narasimhan, Karthik and Press, Ofir},
  booktitle = {NeurIPS},
  year      = {2024},
  note      = {Origin of the bounded ACI pattern adopted by ATLAS.
               Tool-design principles: narrow surface, mechanical bounds,
               evidence-anchored outputs.}
}

% ─── Context engineering ──────────────────────────────────────────────────

@misc{AnthropicContextEng2025,
  title  = {Context Engineering for LLM Agents},
  author = {{Anthropic}},
  year   = {2025},
  month  = {September},
  note   = {Phase-irrelevant instructions degrade performance.
             Basis for layered-loading architecture across all Eidolons.}
}

% ─── Plan-mode and reasoning ──────────────────────────────────────────────

@misc{Wei2022CoT,
  title  = {Chain-of-Thought Prompting Elicits Reasoning in Large Language Models},
  author = {Wei, Jason and Wang, Xuezhi and Schuurmans, Dale and others},
  year   = {2022},
  note   = {Baseline for FORGE's reasoning methodology (in construction).}
}

@misc{Yao2023ToT,
  title  = {Tree of Thoughts: Deliberate Problem Solving with Large Language Models},
  author = {Yao, Shunyu and Yu, Dian and Zhao, Jeffrey and others},
  year   = {2023},
  note   = {Candidate reasoning structure for FORGE.}
}

% ─── Context compaction & memory ──────────────────────────────────────────

@misc{AgentFold2024,
  title  = {AgentFold: Context Compaction at Agent Phase Boundaries},
  author = {{AgentFold Authors}},
  year   = {2024},
  note   = {ATLAS applies AgentFold at phase boundaries; working memory
             holds IDs + anchors, raw content goes to Memex.}
}

% ─── Production patterns (non-paper references) ───────────────────────────

@online{ClaudeCodeDocs,
  title  = {Claude Code Documentation},
  author = {{Anthropic}},
  url    = {https://docs.claude.com/en/docs/claude-code},
  year   = {2025},
  note   = {Plan Mode, skills, subagents — patterns adopted by ATLAS and
             SPECTRA.}
}

@online{CursorDocs,
  title  = {Cursor Rules Documentation},
  author = {{Cursor}},
  url    = {https://docs.cursor.com},
  year   = {2025},
  note   = {MDC rule files and AGENTS.md — part of EIIS host-wiring targets.}
}

@online{AiderBlog,
  title  = {Aider: AI Pair Programming in Your Terminal},
  author = {Gauthier, Paul},
  url    = {https://aider.chat},
  year   = {2024},
  note   = {File-aware editing and test-driven repair loops — influences
             APIVR-$\Delta$'s Implement/Verify/Reflect structure.}
}

% ─── Ongoing — expand as citations are consolidated from per-Eidolon rationale ─