Unable to reproduce the results in the paper

[ghwmx]

Hello, we have completed feature extraction based on the code in GitHub and run the TSPM model, but our results differ significantly from those in the paper. Can you help me check where there is a problem?
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio/Counting accuracy: 76.99(783/1017)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio/Comparative accuracy: 56.40(335/594)
[2025-04-03 20:13:44,562]-[trainutils.py line:459]:Test Audio accuracy: 69.40(1118/1611)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Visual/Counting accuracy: 77.11(923/1197)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Visual/Location accuracy: 76.65(939/1225)
[2025-04-03 20:13:44,562]-[trainutils.py line:459]:Test Visual accuracy: 76.88(1862/2422)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio-Visual/Existential accuracy: 78.14(772/988)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio-Visual/Counting accuracy: 76.21(964/1265)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio-Visual/Location accuracy: 55.43(510/920)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio-Visual/Comparative accuracy: 62.76(691/1101)
[2025-04-03 20:13:44,562]-[trainutils.py line:456]:Test Audio-Visual/Temporal accuracy: 66.42(546/822)
[2025-04-03 20:13:44,562]-[trainutils.py line:459]:Test Audio-Visual accuracy: 68.35(3483/5096)
[2025-04-03 20:13:44,563]-[trainutils.py line:461]:Test Total avg accuracy: 70.80(6463/9129)

[lik1996]

Same issue from my side.