[zamadatix]

That (thankfully) can't compound, so would never be more than a one time offset. E.g. if you report a score of 60% SWE-bench verified for new model A, dumb A down to score 50%, and report a 20% improvement over A with new model B then it's pretty obvious when your last two model blogposts say 60%.

The only way around this is to never report on the same benchmark versions twice, which they include too many to realistically do every release.

[MichealCodes]

The benchmarks are not typically ongoing, we do not often see comparisons between week 1 and week 8. Sprinkle a bit of training on the benchmarks in and you can ensure higher scores for the next model. A perfect scam loop to keep the people happy until they wise up.

[zamadatix]

> The benchmarks are not typically ongoing, we do not often see comparisons between week 1 and week 8

You don't need to compare "A (Week 1)" to "A (Week 8)" to be able to show "B (Week 1)" is genuinely x% better than "A (Week 1)".

[MichealCodes]

As I said sprinkle a bit of benchmarks polluting the training and you have your loop. Each iteration will be better at benchmarks if that's the goal and that goal/context reinforces.

[zamadatix]

Sprinkling in benchmark training isn't a loop, it's just plain cheating. Regardless, not all of these benchmarks are public and, even with mass collusion across the board, it wouldn't make sense only open weight LLMS have been improving.