dissents and votes to reverse the judgment of conviction and order a new trial, with the following memorandum: In this case, the trial court committed two serious errors in its instructions to the jury: (1) that the “[ejvidence with relation to an alibi should be carefully examined” (see People v McFadden, 100 AD2d 520; People v Costales, 87 AD2d 635; People v Wallace, 87 AD2d 895; People v Reed, 83 AD2d 645; People v Rothaar, 75 AD2d 652; People v Fludd, 68 AD2d 409; cf. People v Victor, 62 NY2d 374) and (2) that if the jurors “minds [were] waivering or the scales [were] even or if you have such a reasonable doubt arising out of the credible evidence * * * the benefit of that doubt must be given to the defendant” (see People v Wade, 99 AD2d 474; People v Ortiz, 92 AD2d 595; People v Melville, 90 AD2d 488).
While it is true that no protest was made pursuant to CPL 470.05 (subd 2), we have generally found such instructional error to be so prejudicial as to warrant review in the interest of justice (e.g., People v McFadden, supra; People v Wade, supra; People v Melville, supra; People v Costales, supra; People v Fludd, supra). I fail to perceive why this case should be an exception.
With due respect, the charge, read as a whole, is confusing, and the jury could very well have been misled, especially absent a balancing identification instruction (People v Schellhammer, 97 AD2d 776; cf. People v Victor, supra). More important, this was an extremely close case and the errors could clearly have made a difference. Although the key complaining witness, Hall, identified defendant in court as one of the attackers, and testified that defendant had a moustache at the time of the crime, she admitted testifying at the Wade hearing that the “short” man had no moustache and claimed that, despite the police complaint report which was prepared from information given by the complainants and specified that the man identified as the defendant had a scarred face, she had never told the police that he had facial scars. Perhaps the “certainty” in identification was *576bolstered by the fact that the prosecutor had told Hall before trial that the individual to be tried had a criminal background.
Review of the record also indicates that defendant and his witnesses presented a plausible alibi defense. Surely there can be no quarrel with the proposition that the credibility of the alibi defense was for the jury to assess (e.g., People v De Tore, 34 NY2d 199, 206-207, cert den sub nom. Wedra v New York, 419 US 1025). It seems to me, however, to beg the question to assert, as does the majority, that the jury disbelieved the alibi. That determination was based upon an erroneous instruction and “the proof of guilt in this case, which turned to a great extent on the jury evaluation of the alibi witnesses on the one hand and the prosecution witnesses on the other, was not so compelling as to render the error in the charge harmless” (People v McFadden, 100 AD2d 520, 521, supra).
I also note that, in his cross-examination of defense witnesses, the prosecutor violated the rules set forth in People v Dawson (50 NY2d 311), decided a full year before the trial in this case. Without the requisite bench conference and over apparent objection, the prosecutor questioned these witnesses on their failure to go to the police with exculpatory information (see People v Schellhammer, 97 AD2d 776, supra; People v Muniz, 89 AD2d 611). The error was compounded by the prosecutor’s comments in summation and by the trial court’s failure to give an appropriate limiting instruction.
In sum, defendant’s trial was tainted by prejudicial legal error. Accordingly, I dissent and cast my vote for reversal.