Files changed (1) hide show
  1. README.md +195 -4
README.md CHANGED
@@ -1,9 +1,200 @@
1
  ---
2
  license: apache-2.0
3
  ---
4
- Our D-CORE paper: arxiv.org/abs/2602.02160
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
 
6
  ### ACEBench
7
- | Model | bool | enum | number | list | object_short | object_deep | atom | single_turn_singal_function | single_turn_parallel_function | single_turn | multiple_turn_switch | multiple_turn_adjust | multiple_turn | similar_api | profile | normal_summary | incomplete | error | irrelevant | special_summary | agent_multi_turn | agent_multi_turn_process | agent_multi_step | agent_multi_step_process | agent_summary | Summary |
8
- |:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|:---|
9
- |**D-CORE-8B**| 0.9 | 0.98 | 0.98 | 0.98 | 0.36 | 0.76 | 0.827 | 0.85 | 0.7 | 0.775 | 0.64 | 0.6 | 0.62 | 0.78 | 0.82 | 0.779 | 0.58 | 0.82 | 0.96 | 0.787 | 0.433 | 0.668 | 0.75 | 0.808 | 0.592 | 0.752 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
  ---
4
+ <!-- Our D-CORE paper: arxiv.org/abs/2602.02160 -->
5
+ [Paper](arxiv.org/abs/2602.02160) | [Code](https://github.com/alibaba/EfficientAI) | Dataset(Comming soon)
6
+
7
+ ### BFCL
8
+ In our network environment, for the Web Search No Snippet task, we are unable to access certain websites (e.g., Wikipedia), which results in some deviation in the No Snippet scores.
9
+
10
+ <table>
11
+ <thead>
12
+ <tr>
13
+ <th rowspan="4">Model</th>
14
+ <th rowspan="4">Overall</th>
15
+ <th colspan="7" style="text-align:center">Agentic</th>
16
+ <th colspan="5" style="text-align:center">Multi Turn</th>
17
+ <th colspan="10" style="text-align:center">Single Turn</th>
18
+ <th colspan="2" style="text-align:center">Hallucination Measurement</th>
19
+ <th colspan="2" style="text-align:center">Format Sensitivity</th>
20
+ </tr>
21
+ <tr>
22
+ <th colspan="3" style="text-align:center">Web Search</th>
23
+ <th colspan="4" style="text-align:center">Memory</th>
24
+ <th rowspan="3">Overall Acc</th>
25
+ <th rowspan="3">Base</th>
26
+ <th rowspan="3">Miss Func</th>
27
+ <th rowspan="3">Miss Param</th>
28
+ <th rowspan="3">Long Context</th>
29
+ <th colspan="5" style="text-align:center">Non-live</th>
30
+ <th colspan="5" style="text-align:center">Live</th>
31
+ <th rowspan="3">Relevance</th>
32
+ <th rowspan="3">Irrelevance</th>
33
+ <th rowspan="3">Max Delta</th>
34
+ <th rowspan="3">SD</th>
35
+ </tr>
36
+ <tr>
37
+ <th rowspan="2">Summary</th>
38
+ <th rowspan="2">Base</th>
39
+ <th rowspan="2">No Snippet</th>
40
+ <th rowspan="2">Summary</th>
41
+ <th rowspan="2">KV</th>
42
+ <th rowspan="2">Vector</th>
43
+ <th rowspan="2">Recusive Sum</th>
44
+ <th rowspan="2">Overall Acc</th>
45
+ <th rowspan="2">Simple</th>
46
+ <th rowspan="2">Multiple</th>
47
+ <th rowspan="2">Parallel</th>
48
+ <th rowspan="2">Multiple Parallel</th>
49
+ <th rowspan="2">Overall Acc</th>
50
+ <th rowspan="2">Simple</th>
51
+ <th rowspan="2">Multiple</th>
52
+ <th rowspan="2">Parallel</th>
53
+ <th rowspan="2">Multiple Parallel</th>
54
+ </tr>
55
+ <tr>
56
+ </tr>
57
+ </thead>
58
+ <tbody>
59
+ <tr>
60
+ <td style="white-space: nowrap"><strong>D-CORE-8B</strong></td>
61
+ <td style="text-align: center">53.15</td>
62
+ <td style="text-align: center">23.00</td>
63
+ <td style="text-align: center">36.00</td>
64
+ <td style="text-align: center">10.00</td>
65
+ <td style="text-align: center">19.14</td>
66
+ <td style="text-align: center">9.03</td>
67
+ <td style="text-align: center">16.77</td>
68
+ <td style="text-align: center">31.61</td>
69
+ <td style="text-align: center">64.88</td>
70
+ <td style="text-align: center">75.50</td>
71
+ <td style="text-align: center">65.00</td>
72
+ <td style="text-align: center">60.50</td>
73
+ <td style="text-align: center">58.50</td>
74
+ <td style="text-align: center">86.85</td>
75
+ <td style="text-align: center">75.92</td>
76
+ <td style="text-align: center">92.50</td>
77
+ <td style="text-align: center">92.00</td>
78
+ <td style="text-align: center">87.00</td>
79
+ <td style="text-align: center">75.80</td>
80
+ <td style="text-align: center">78.29</td>
81
+ <td style="text-align: center">75.02</td>
82
+ <td style="text-align: center">100.00</td>
83
+ <td style="text-align: center">66.67</td>
84
+ <td style="text-align: center">75.00</td>
85
+ <td style="text-align: center">89.99</td>
86
+ <td style="text-align: center">75.0</td>
87
+ <td style="text-align: center">24.67</td>
88
+ </tr>
89
+ </tbody>
90
+ </table>
91
+
92
+ ### Tau-Bench & Tau2-Bench
93
+ We use Qwen3-235B-A22B-Instruct-2507 as the user model. For each task, we sample 5 times and take the average as the final result.
94
+ <table>
95
+ <thead>
96
+ <tr>
97
+ <th rowspan="3">Model</th>
98
+ <th colspan="3" style="text-align:center">Tau-Bench</th>
99
+ <th colspan="4" style="text-align:center">Tau2-Bench</th>
100
+ </tr>
101
+ <tr>
102
+ <th rowspan="2">Overall</th>
103
+ <th rowspan="2">Retail</th>
104
+ <th rowspan="2">Airline</th>
105
+ <th rowspan="2">Overall</th>
106
+ <th rowspan="2">Retail</th>
107
+ <th rowspan="2">Airline</th>
108
+ <th rowspan="2">Telecom</th>
109
+ </tr>
110
+ <tr>
111
+ </tr>
112
+ </thead>
113
+ <tbody>
114
+ <tr>
115
+ <td style="white-space: nowrap"><strong>D-CORE-8B</strong></td>
116
+ <td style="text-align: center">44.9</td>
117
+ <td style="text-align: center">53.0</td>
118
+ <td style="text-align: center">36.8</td>
119
+ <td style="text-align: center">35.8</td>
120
+ <td style="text-align: center">43.2</td>
121
+ <td style="text-align: center">37.1</td>
122
+ <td style="text-align: center">27.2</td>
123
+ </tr>
124
+ </tbody>
125
+ </table>
126
 
127
  ### ACEBench
128
+ <table>
129
+ <thead>
130
+ <tr>
131
+ <th rowspan="3">Model</th>
132
+ <th rowspan="3">Overall</th>
133
+ <th colspan="7" style="text-align:center">Atom</th>
134
+ <th colspan="3" style="text-align:center">Single Turn</th>
135
+ <th colspan="3" style="text-align:center">Multi Turn</th>
136
+ <th rowspan="3" style="text-align:center">Similar API</th>
137
+ <th rowspan="3" style="text-align:center">Preference</th>
138
+ <th rowspan="3" style="text-align:center">Summary</th>
139
+ <th colspan="4" style="text-align:center">Special</th>
140
+ <th colspan="5" style="text-align:center">Agent</th>
141
+ </tr>
142
+ <tr>
143
+ <th rowspan="2">Summary</th>
144
+ <th rowspan="2">Bool</th>
145
+ <th rowspan="2">Enum</th>
146
+ <th rowspan="2">Number</th>
147
+ <th rowspan="2">List</th>
148
+ <th rowspan="2">Object Short</th>
149
+ <th rowspan="2">Object Deep</th>
150
+ <th rowspan="2">Summary</th>
151
+ <th rowspan="2">Singal Function</th>
152
+ <th rowspan="2">Parallel Function</th>
153
+ <th rowspan="2">Summary</th>
154
+ <th rowspan="2">Switch</th>
155
+ <th rowspan="2">Adjust</th>
156
+ <th rowspan="2">Summary</th>
157
+ <th rowspan="2">Incomplete</th>
158
+ <th rowspan="2">Error</th>
159
+ <th rowspan="2">Irrelevant</th>
160
+ <th rowspan="2">Summary</th>
161
+ <th rowspan="2">Multi Turn</th>
162
+ <th rowspan="2">Multi Turn Process</th>
163
+ <th rowspan="2">Multi Step</th>
164
+ <th rowspan="2">Multi Step Process</th>
165
+ </tr>
166
+ <tr>
167
+ </tr>
168
+ </thead>
169
+ <tbody>
170
+ <tr>
171
+ <td style="white-space: nowrap"><strong>D-CORE-8B</strong></td>
172
+ <td style="text-align: center">75.2</td>
173
+ <td style="text-align: center">82.7</td>
174
+ <td style="text-align: center">90.0</td>
175
+ <td style="text-align: center">98.0</td>
176
+ <td style="text-align: center">98.0</td>
177
+ <td style="text-align: center">98.0</td>
178
+ <td style="text-align: center">36.0</td>
179
+ <td style="text-align: center">76.0</td>
180
+ <td style="text-align: center">77.5</td>
181
+ <td style="text-align: center">85.0</td>
182
+ <td style="text-align: center">70.0</td>
183
+ <td style="text-align: center">62.0</td>
184
+ <td style="text-align: center">64.0</td>
185
+ <td style="text-align: center">60.0</td>
186
+ <td style="text-align: center">78.0</td>
187
+ <td style="text-align: center">82.0</td>
188
+ <td style="text-align: center">77.9</td>
189
+ <td style="text-align: center">78.7</td>
190
+ <td style="text-align: center">58.0</td>
191
+ <td style="text-align: center">82.0</td>
192
+ <td style="text-align: center">96.0</td>
193
+ <td style="text-align: center">59.2</td>
194
+ <td style="text-align: center">43.3</td>
195
+ <td style="text-align: center">66.8</td>
196
+ <td style="text-align: center">75.0</td>
197
+ <td style="text-align: center">80.8</td>
198
+ </tr>
199
+ </tbody>
200
+ </table>