Hello Ovis Team,
I have a question regarding the performance benchmarks reported in the technical report.
In Table 2 & 3, Ovis2.5-9B achieves an impressive average score of 78.3 on the OpenCompass suite. I noticed that some comparison models in the same table are explicitly labeled with "Thinking" (e.g., GLM-4.1V-9B-Thinking), whereas Ovis2.5-9B is not.
Could you please clarify whether the reported score of 78.3 was achieved using the standard inference mode or with the optional "Thinking mode" enabled?
Hello Ovis Team,
I have a question regarding the performance benchmarks reported in the technical report.
In Table 2 & 3, Ovis2.5-9B achieves an impressive average score of 78.3 on the OpenCompass suite. I noticed that some comparison models in the same table are explicitly labeled with "Thinking" (e.g., GLM-4.1V-9B-Thinking), whereas Ovis2.5-9B is not.
Could you please clarify whether the reported score of 78.3 was achieved using the standard inference mode or with the optional "Thinking mode" enabled?