tesraghavan kirito011024 commited on
Commit
6c74d50
·
0 Parent(s):

Duplicate from Kirito-Lab/EntCollabBench

Browse files

Co-authored-by: Tao Yu <kirito011024@users.noreply.huggingface.co>

.gitattributes ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.avro filter=lfs diff=lfs merge=lfs -text
4
+ *.bin filter=lfs diff=lfs merge=lfs -text
5
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
6
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
7
+ *.ftz filter=lfs diff=lfs merge=lfs -text
8
+ *.gz filter=lfs diff=lfs merge=lfs -text
9
+ *.h5 filter=lfs diff=lfs merge=lfs -text
10
+ *.joblib filter=lfs diff=lfs merge=lfs -text
11
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
12
+ *.lz4 filter=lfs diff=lfs merge=lfs -text
13
+ *.mds filter=lfs diff=lfs merge=lfs -text
14
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
15
+ *.model filter=lfs diff=lfs merge=lfs -text
16
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
17
+ *.npy filter=lfs diff=lfs merge=lfs -text
18
+ *.npz filter=lfs diff=lfs merge=lfs -text
19
+ *.onnx filter=lfs diff=lfs merge=lfs -text
20
+ *.ot filter=lfs diff=lfs merge=lfs -text
21
+ *.parquet filter=lfs diff=lfs merge=lfs -text
22
+ *.pb filter=lfs diff=lfs merge=lfs -text
23
+ *.pickle filter=lfs diff=lfs merge=lfs -text
24
+ *.pkl filter=lfs diff=lfs merge=lfs -text
25
+ *.pt filter=lfs diff=lfs merge=lfs -text
26
+ *.pth filter=lfs diff=lfs merge=lfs -text
27
+ *.rar filter=lfs diff=lfs merge=lfs -text
28
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
29
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
30
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
31
+ *.tar filter=lfs diff=lfs merge=lfs -text
32
+ *.tflite filter=lfs diff=lfs merge=lfs -text
33
+ *.tgz filter=lfs diff=lfs merge=lfs -text
34
+ *.wasm filter=lfs diff=lfs merge=lfs -text
35
+ *.xz filter=lfs diff=lfs merge=lfs -text
36
+ *.zip filter=lfs diff=lfs merge=lfs -text
37
+ *.zst filter=lfs diff=lfs merge=lfs -text
38
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
39
+ # Audio files - uncompressed
40
+ *.pcm filter=lfs diff=lfs merge=lfs -text
41
+ *.sam filter=lfs diff=lfs merge=lfs -text
42
+ *.raw filter=lfs diff=lfs merge=lfs -text
43
+ # Audio files - compressed
44
+ *.aac filter=lfs diff=lfs merge=lfs -text
45
+ *.flac filter=lfs diff=lfs merge=lfs -text
46
+ *.mp3 filter=lfs diff=lfs merge=lfs -text
47
+ *.ogg filter=lfs diff=lfs merge=lfs -text
48
+ *.wav filter=lfs diff=lfs merge=lfs -text
49
+ # Image files - uncompressed
50
+ *.bmp filter=lfs diff=lfs merge=lfs -text
51
+ *.gif filter=lfs diff=lfs merge=lfs -text
52
+ *.png filter=lfs diff=lfs merge=lfs -text
53
+ *.tiff filter=lfs diff=lfs merge=lfs -text
54
+ # Image files - compressed
55
+ *.jpg filter=lfs diff=lfs merge=lfs -text
56
+ *.jpeg filter=lfs diff=lfs merge=lfs -text
57
+ *.webp filter=lfs diff=lfs merge=lfs -text
58
+ # Video files - compressed
59
+ *.mp4 filter=lfs diff=lfs merge=lfs -text
60
+ *.webm filter=lfs diff=lfs merge=lfs -text
61
+ overview.pdf filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,156 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ configs:
3
+ - config_name: approval_multi_tasks
4
+ data_files: "approval_multi_task_20.json"
5
+ - config_name: approval_tasks
6
+ data_files: "approval_tasks_80.json"
7
+ - config_name: mcp_multi_tasks
8
+ data_files: "mcp_multi_tasks_40.json"
9
+ - config_name: mcp_tasks
10
+ data_files: "mcp_tasks_160.json"
11
+ ---
12
+
13
+ # EntCollabBench
14
+
15
+ EntCollabBench is a benchmark dataset for evaluating **enterprise-oriented multi-agent collaboration**. It focuses on realistic workplace task execution where an entry agent must complete user requests by using tools, interacting with enterprise systems, and delegating subtasks to specialized peer agents when needed.
16
+
17
+ The dataset is designed for benchmarking agent systems that operate in a simulated enterprise environment with role separation, service boundaries, approval workflows, and observable side effects.
18
+
19
+ ![EntCollabBench overview](./overview.png)
20
+
21
+
22
+ ## Dataset Summary
23
+
24
+ EntCollabBench contains four task files and two supporting resource archives:
25
+
26
+ - `mcp_tasks_160.json`: 160 single-task samples centered on enterprise MCP tool use.
27
+ - `mcp_multi_tasks_40.json`: 40 multi-step or multi-subtask collaboration samples.
28
+ - `approval_tasks_80.json`: 80 approval-oriented tasks requiring document-grounded decisions.
29
+ - `approval_multi_task_20.json`: 20 multi-subtask approval and coordination tasks.
30
+ - `seed.zip`: seed data used to initialize the simulated enterprise services.
31
+ - `local_data.zip`: local approval documents and policy materials used by approval agents.
32
+
33
+ In total, the benchmark includes **300 tasks**:
34
+
35
+ - 200 MCP-oriented tasks
36
+ - 100 approval-oriented tasks
37
+
38
+ Depending on file format, one task batch may contain either:
39
+
40
+ - a single task, or
41
+ - multiple subtasks grouped under one `task_id`
42
+
43
+ ## Supported Research Uses
44
+
45
+ This dataset is intended for research on:
46
+
47
+ - multi-agent task decomposition and delegation
48
+ - tool-use in structured enterprise environments
49
+ - cross-role collaboration under access constraints
50
+ - approval reasoning grounded in local policy documents
51
+ - trajectory-based evaluation of agent workflows
52
+ - benchmarking end-to-end operational success, not just final text quality
53
+
54
+ ## Data Fields
55
+
56
+ Common fields appearing across files include:
57
+
58
+ - `task_id`: unique identifier for a task batch
59
+ - `description`: optional natural language description of the batch
60
+ - `task`: task instruction for single-task samples
61
+ - `target_agent`: initial agent for single-task samples
62
+ - `sub_task_list`: list of subtasks in multi-task samples
63
+ - `sub_task_id`: unique identifier for a subtask within a batch
64
+ - `user_prompt`: user-facing instruction for a subtask
65
+ - `beginning_agent`: initial agent assigned to the subtask
66
+ - `ground_truth`: optional structured reference actions or expected tool-use path
67
+ - `state_export`: optional specification for evaluating state changes in enterprise services
68
+
69
+ The benchmark code normalizes several legacy aliases, including:
70
+
71
+ - `task` or `user_prompt`
72
+ - `target_agent`, `begin_agent`, or `beginning_agent`
73
+ - `sub_task_list` or `task_list`
74
+
75
+ ## Agent Roles
76
+
77
+ The benchmark environment uses specialized enterprise agents, including:
78
+
79
+ - `it_service_desk_l1`
80
+ - `it_change_engineer`
81
+ - `hr_service_specialist`
82
+ - `customer_support_specialist`
83
+ - `knowledge_base_specialist`
84
+ - `collaboration_ops_specialist`
85
+ - `developer_engineer`
86
+ - `qa_test_engineer`
87
+ - `finance_approval_specialist`
88
+ - `legal_approval_specialist`
89
+ - `procurement_approval_specialist`
90
+
91
+ These agents operate under role-specific tool and responsibility boundaries.
92
+
93
+ ## Enterprise Service Domains
94
+
95
+ Tasks may involve one or more simulated enterprise systems, including:
96
+
97
+ - `calendar`
98
+ - `csm`
99
+ - `drive`
100
+ - `email`
101
+ - `gitea`
102
+ - `hr`
103
+ - `itsm`
104
+ - `teams`
105
+
106
+ Approval tasks additionally rely on local policy and submission documents supplied through `local_data.zip`.
107
+
108
+ ## Example Use
109
+
110
+ ### Load a task file
111
+
112
+ ```python
113
+ import json
114
+
115
+ with open("mcp_tasks_160.json", "r", encoding="utf-8") as f:
116
+ data = json.load(f)
117
+
118
+ print(type(data))
119
+ print(data[0] if isinstance(data, list) else data)
120
+ ```
121
+
122
+ ## Additional Resources
123
+
124
+ ### `seed.zip`
125
+
126
+ Contains seed artifacts used to initialize the simulated enterprise backend services. These resources are required for reproducing the interactive environment used by the benchmark.
127
+
128
+ ### `local_data.zip`
129
+
130
+ Contains local rulebooks and policy documents for approval specialists, such as finance, legal, and procurement approval materials.
131
+
132
+ Representative document categories include:
133
+
134
+ - finance policies
135
+ - legal and privacy policies
136
+ - procurement and vendor policies
137
+
138
+ These materials are used as local workspace evidence during approval evaluation.
139
+
140
+ ## Citation
141
+
142
+ If you use EntCollabBench in your research, please cite:
143
+
144
+ ```bibtex
145
+ @misc{yu2026allinoneagentbenchmarkingrolespecialized,
146
+ title={Beyond the All-in-One Agent: Benchmarking Role-Specialized Multi-Agent Collaboration in Enterprise Workflows},
147
+ author={Tao Yu and Hao Wang and Changyu Li and Shenghua Chai and Minghui Zhang and Zhongtian Luo and Yuxuan Zhou and Haopeng Jin and Zhaolu Kang and Jiabing Yang and YiFan Zhang and Xinming Wang and Hongzhu Yi and Zheqi He and Jing-Shu Zheng and Xi Yang and Yan Huang and Liang Wang},
148
+ year={2026},
149
+ eprint={2605.08761},
150
+ archivePrefix={arXiv},
151
+ primaryClass={cs.MA},
152
+ url={https://arxiv.org/abs/2605.08761},
153
+ }
154
+ ```
155
+
156
+
approval_multi_task_20.json ADDED
The diff for this file is too large to render. See raw diff
 
approval_tasks_80.json ADDED
The diff for this file is too large to render. See raw diff
 
local_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb9ee705bf609133d25de60f07af795bed14a0c1f57b08b5bb91cd03f78fe48e
3
+ size 1459558
mcp_multi_tasks_40.json ADDED
The diff for this file is too large to render. See raw diff
 
mcp_tasks_160.json ADDED
The diff for this file is too large to render. See raw diff
 
overview.png ADDED

Git LFS Details

  • SHA256: 0de1374682486f052f124232155b3b9d79d8c82210205f42c0fbf30cdab07d55
  • Pointer size: 131 Bytes
  • Size of remote file: 354 kB
seed.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ef0f60f9b125769639264075bfd586c8d4b2eda68e7971b58c364a16871b8ec
3
+ size 1629329