Algorithm 1: Anchor-Teeth Guided Inference Algorithm.

Input: Sorted predicted teeth list P with class labels and probabilities Output: Refined teeth list with corrected class labels Define reference rule: R = [8,7,6,5,4,3,2,1,1,2,3,4,5,6,7,8] Extract predicted labels Lp and probabilities Sp from P Initialize: max_len = 0 best_pred_start = −1 best_ref_start = −1 best_score = −1 For each r in [0, |R|): For each p in [0, |Lp|): ⚬ match_len = 0, score = 0 ⚬ While R[r + match_len] == Lp[p + match_len]: ▪ score += Sp[p + match_len] ▪ match_len += 1 ⚬ If (match_len > max_len) OR (match_len == max_len AND score > best_score): ▪ Update best_pred_start, best_ref_start ▪ max_len = match_len ▪ best_score = score If max_len > 0 For each index i in P: ⚬ ref_idx = best_ref_start + (i − best_pred_start) ⚬ If 0 ≤ ref_idx < |R|: ▪ P[i].cls = R[ref_idx] Return P