Reward hacking is swamping model intelligence gains

This story was filed as a headline only — the news service holds no English full text for it. Read the original at Cursor Blog →