dsh-llm-as-a-verifier

English | 中文

Brings LLM-as-a-Verifier — the unified verification framework — into DeepSeek Harness (dsh) as a tool plugin. The agent can now verify its own candidates with fine-grained probabilistic feedback: instead of a binary good/bad judgement, the verifier's full logprob distribution over a 20-point score scale is read and its expectation taken.

What you get

Three model-facing tools, registered automatically after install:

Why finer than LLM-as-a-Judge? The upstream framework's key ideas are: ① fine-grained scoring granularity (20-letter scale); ② expectation over the full logprob distribution of score tokens; ③ reliability via repeated evaluations and criteria decomposition. Building on its open-source implementation, this plugin adapts score extraction, pairwise prompts, the pivot tournament, and progress tracking, adds a logprob backend and token accounting, and packages them as a Cordis tool plugin for DSH.

Install

dsh plugin --profile web add dsh-llm-as-a-verifier

Requires dsh ≥ 0.1.0-rc.6 and Node ≥ 18. Restart dsh web (or wait for HMR).

Configure the verifier backend

The verifier model must be an OpenAI-compatible service returning token-level logprobs: DeepSeek's hosted API, a local vLLM/SGLang server, OpenAI, etc.

In your profile config (~/.dsh/profiles/<name>/cordis.patch.yml or ~/.dsh/cordis.patch.yml):

- id: llm-verifier
  config:
    baseUrl: https://api.deepseek.com   # or vLLM: http://localhost:8000/v1
    apiKey: '${DEEPSEEK_API_KEY}'       # environment variables preferred
    model: deepseek-v4-flash            # omitted: DeepSeek defaults to deepseek-v4-flash, others auto-probe /models
    maxConcurrency: 8

Credential resolution order (upstream parity): plugin config → OPENAI_BASE_URL + OPENAI_API_KEY → DEEPSEEK_API_KEY (implies the DeepSeek endpoint with thinking enabled). With no credentials configured, tool registration still works and calls fail with MissingAPIKeyError only when executed.

export DEEPSEEK_API_KEY=sk-...   # the simplest setup

Usage

Once installed, just ask the agent in the conversation:

I wrote three candidate implementations. Use verify_select with
"correctness" and "performance" criteria to pick the best one, then use
verify_track to check whether my earlier fix steps made progress.

Configuration reference

Tool arguments (nEvaluations, pivots, seed, groundTruthNote, ...) mirror the upstream llm_verifier Python package; see the user guide and SOP.

Library use

import { Verifier } from 'dsh-llm-as-a-verifier'

const verifier = new Verifier({ baseUrl: 'http://localhost:8000/v1' })
const { scoreA, scoreB } = await verifier.compare(problem, a, b, { Correctness: '...' })
const result = await verifier.select(problem, candidates, { Correctness: '...' }, { pivots: 2 })
const curve = await verifier.track(problem, steps, { checkpoints: [1, 3] })

Development

npm ci
npm run check   # typecheck + vitest (76 cases incl. end-to-end against a local mock logprobs server)
npm run build

License

This project is released under the MIT License.

Acknowledgments

Portions of the scoring and verification implementation are adapted from LLM-as-a-Verifier. Thanks to its authors and contributors; the relevant copyright and license notices are preserved in LICENSE.