Ubicloud: vLLM 0.31's constrained decoding took Kimi K2.6 tool-call schema accuracy from 92% to 100% in its tests
Ubicloud reports that when serving Kimi K2.6 with vLLM 0.28, the expected tool-call schema was produced only 92% of the time, meaning about 2 in 25 tool calls failed. vLLM recently added constrained decoding, a technique to guarantee schema adherence, released as a server-side flag in vLLM 0.31. The authors state that enabling it yields 100% schema accuracy with newer Kimi and DeepSeek models. The post explains tool-call schemas and why schema accuracy matters for agentic workloads, using Moonshot AI's K2VV ToolCall benchmark as the reference.
One primary source; other links are context, not independent corroboration.
- Primary sourceHacker News discovery · source check open
- Discovery linkHacker News
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
Open full article view →- Linked article+30
- Fetched source page+30
- No editor link review+0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
Ubicloud reports that when serving Kimi K2.6 with vLLM 0.28, the expected tool-call schema was produced only 92% of the time, meaning about 2 in 25 tool calls failed. vLLM recently added constrained decoding, a technique to guarantee schema adherence, released as a server-side flag in vLLM 0.31. The authors state that enabling it yields 100% schema accuracy with newer Kimi and DeepSeek models. The post explains tool-call schemas and why schema accuracy matters for agentic workloads, using Moonshot AI's K2VV ToolCall benchmark as the reference.
Key findings from the stored brief
- With vLLM 0.28, Kimi K2.6 produced the expected tool-call schema only 92% of the time.
- vLLM 0.31 exposes constrained decoding, which guarantees schema adherence, as a server-side flag.
- The authors report 100% schema accuracy with newer Kimi and DeepSeek models when the flag is enabled.
- Schema accuracy is measured with Moonshot AI's K2VV ToolCall benchmark of 2000 requests.
Source claims not independently checked; the 100% figure comes from the authors' own tests.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
A publication revision history is not stored for this signal. New engagement snapshots do not mean the source itself changed.
Why this may matter
Malformed tool calls waste agent retries, pollute context, and degrade multi-turn agentic workloads. A reported path to guaranteed schema adherence on open-weight models could make self-hosted agent serving more reliable.
Limits and open questions
Source claims not independently checked; the 100% figure comes from the authors' own tests.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Discussion
Ask about the source, add context, or point readers to a relevant follow-up.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.