[faresobeid]

renderers.client.generate() passes two more optional prime-rl /inference/v1/generate fields through. Both are pure pass-through: no request-side change, no parsing or validation (same as sampling_mask). Stock vLLM sends neither, and then the keys are None.

• sampling_mask_logprobs = choice.get("sampling_mask_logprobs"): the sampler's logprobs at the sampling-mask ids, packed base64 float32 {data, shape, dtype}, parallel to the packed sampling_mask ids. verifiers decodes it (feat(v1): optional sampler logprobs on SamplingMask verifiers#2756). Consumer: prime-rl score centering, feat: exact score centering over the sampling mask (arXiv:2609.20807) prime-rl#3794.
• payload = choice.get("payload"): by-handle segments of a shared file that prime-rl's inference server writes instead of returning routed_experts / sampling_mask inline. verifiers puts them on traces (feat(v1): by-handle payload segments on traces verifiers#2765). Consumer: feat: pass router-replay ids and sampling masks by handle on multi-node runs prime-rl#3902.

Order

This PR with PrimeIntellect-ai/verifiers#2756 (-> prime-rl#3794) and PrimeIntellect-ai/verifiers#2765 (-> prime-rl#3902).

Validation

• pytest tests/test_client.py --noconftest (32 passed), ruff check, ruff format --check on renderers/client.py.

Note

Pass through sampling-mask logprobs and by-handle payload segments in generation client

The async generation client in client.py now reads sampling-mask logprobs from the response choice and includes them in the result. It also forwards the choice's payload field for by-handle side arrays such as routed experts or sampling masks. Missing fields are returned as None.

^{Macroscope summarized 196f97e.}

---

Note

Low Risk
Backward-compatible optional response pass-through with no request, validation, or parsing logic changes.

Overview
generate() now surfaces two optional prime-rl /inference/v1/generate choice fields on its result dict, with no request or parsing changes (same pattern as existing sampling_mask / routed_experts pass-through).

sampling_mask_logprobs carries packed sampler logprobs aligned with the sampling-mask token ids when score centering is enabled on the inference server. payload forwards by-handle segments pointing at a shared sidecar file when the server offloads large routed_experts / sampling_mask blobs instead of inlining them. Engines that omit either field leave the keys as None.

Docstrings and inline comments were updated to document the new return keys.

^{Reviewed by Cursor Bugbot for commit 196f97e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Needs Human Approval

Macroscope's review found this PR requires human approval — The patch is a small, backward-compatible addition that exposes optional response metadata without changing request behavior. Two unresolved substantive comments identify potential reasoning/parsing correctness issues elsewhere in the PR, so human review is warranted.

Notes:

• Macroscope's correctness review did not run, so approvability was decided on eligibility alone.

No code changes detected at 196f97e. Prior analysis still applies.

^{You can add or adjust custom eligibility rules. Learn more.}

Dismissing prior approval to re-evaluate d3a1af0

Dismissing prior approval to re-evaluate 030b208

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 030b208. Configure here.}

[cursor]

Reasoning text drops think markers

Medium Severity

scan_reasoning already skips the leading <think> opener, then deletes every later open_id and every <think> substring before decoding what remains. Reasoning that mentions the tag, or a truncated </think>, loses those characters, and neighboring tokens can merge into different text.

 

^{Reviewed by Cursor Bugbot for commit 030b208. Configure here.}

[cursor]

DeepSeek reasoning boundary whitespace stripped

Medium Severity

parse_deepseek_v4 always rstrips reasoning and turns whitespace-only content into "". The newline the model emits before </think>, and a content channel that is only that separator, disappear from the parsed message.

 

^{Reviewed by Cursor Bugbot for commit 030b208. Configure here.}