test: add regression test for eval sample logging when sample is a list

[devin-ai-integration]

Summary

Adds a regression test for the bug fixed in 8efb116 ([fix] eval sample logging when sample is a list (#1836)).

The bug: eval_rollout_single_dataset accessed .prompt / .response / .reward directly on the sample variable, which could be a list[Sample] (when n_samples_per_prompt > 1), causing an AttributeError.

The fix extracted logged_sample = sample[0] if isinstance(sample, list) else sample before logging.

This PR adds three focused unit tests in tests/utils/test_eval_sample_logging.py:

1. test_eval_sample_logging_with_list_sample — verifies that when sample is a list, the first element's prompt, response, and reward are correctly logged
2. test_eval_sample_logging_with_single_sample — verifies the non-list (single Sample) path still works
3. test_eval_sample_logging_list_raises_without_fix — proves the original bug existed by showing list.prompt raises AttributeError

Review & Testing Checklist for Human

• Verify the helper _log_eval_sample mirrors the fixed logging logic in eval_rollout_single_dataset (line 576-581 of slime/rollout/sglang_rollout.py)
• Run pytest tests/utils/test_eval_sample_logging.py -v to confirm all 3 tests pass

Notes

The tests are lightweight unit tests that only depend on slime.utils.types.Sample and logging — no GPU or external services required.

Link to Devin session: https://app.devin.ai/sessions/f4670acf0c004d6691b0aa0b83a07b74
Requested by: @ShaneGZhu

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring