The following column format is used in the respective stats.txt files generated for all programs:
Column  1: sequence pair identifier
Column  2: null_model message length 
Column  3: alignment_model message length (best alignment hypothesis)
Column  4: compression (null_model - alignment_model message lengths. Positive: Accept; Negative: Reject)
Column  5: I(A) -- Alignment Complexity
Column  6: I(<S,T>|A) -- Fidelity of sequence pair <S,T> given the alignment relationship
Column  7: optimized sequence-distance parameter n 
Column  8: |S| -- length of S
Column  9: |T| -- length of T
Column 10: |A| -- length of A
Column 11: nMatches in A
Column 12: nIdentical amino acid pairs in A
Column 13: %-identity of A

For MML based runs, the subsequent columns after '|' symbol indicate:
Column 14: I_marginal(<S,T>) -- negative log marginal probability
Column 15: compression (null_model - I_marginal) 
Column 16: optimized sequence-distance parameter n -- over marginal probability inference