fix: raise kimi-k3 token budget to 8k (reasoning tokens exhausted 3k on hard prompts)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

scripts/collect_responses.py

@@ -61,7 +61,7 @@ def run_frontier_pass(items: list[dict], out: Path, model: str = KIMI_MODEL, max
             # at 1000 tokens, harder prompts exhausted the budget mid-reasoning and
             # returned empty content (finish_reason="length", content=None). 3000
             # leaves enough room for reasoning + a full visible answer.
-            text, usage = chat(client, model, [{"role": "user", "content": it["prompt"]}], max_tokens=3000)
+            text, usage = chat(client, model, [{"role": "user", "content": it["prompt"]}], max_tokens=8000)
             if not text or not text.strip():
                 raise ValueError("empty response content (likely reasoning-token budget exhaustion)")
         except Exception as e: