wenmin-wu/ds-skills/mbr-decoding-reranking

Claude Code and agent skill by wenmin-wu/ds-skills

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Minimum Bayes Risk decoding — select the candidate with highest average chrF++ agreement against all others in the pool

Install

npx vskill@latest install wenmin-wu/ds-skills/mbr-decoding-reranking

SKILL.md

The same SKILL.md was published earlier as majiayu000/claude-skill-registry-data/mbr-decoding-reranking.

MBR Decoding Reranking

Overview

Instead of picking the single highest-probability beam, generate a pool of candidates and select the one most agreed-upon by the others. Uses chrF++ (character F-score with word bigrams) as the utility metric. Consistently outperforms pure beam search for translation tasks.

Quick Start

python
import sacrebleu
import numpy as np
def mbr_select(candidates, pool_cap=32):
    metric = sacrebleu.metrics.CHRF(word_order=2)
    unique = list(dict.fromkeys(c.strip() for c in candidates if c.strip()))
    pool = unique[:pool_cap]
    n = len(pool)
    if n <= 1:
        return pool[0] if pool else ""
    scores = np.zeros(n)
    for i in range(n):
        for j in range(n):
            if i != j:
                scores[i] += metric.sentence_score(pool[i], [pool[j]]).score
        scores[i] /= (n - 1)
    return pool[int(np.argmax(scores))]

Key Decisions

  • chrF++ over BLEU: more robust at sentence level, handles morphologically rich languages better
  • Pool cap: 32 candidates balances quality vs O(n^2) pairwise cost
  • Deduplicate first: removes exact duplicates before scoring to avoid self-reinforcing bias
  • Combine with multi-temperature sampling: MBR needs diverse candidates to work well

References

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

See the full security report

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

See evaluation results

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/wenmin-wu/ds-skills/mbr-decoding-reranking/badge)](https://verified-skill.com/skills/wenmin-wu/ds-skills/mbr-decoding-reranking)

Something wrong with this skill or its page? Tell us.