AI/ML●●Solid
verbatimeter - check how grounded your LLM / RAG agent is in real-time
Deterministic groundedness verification using longest common subsequence instead of judge models or embeddings.
Big BrainNiche Gem
pobonin
103d ago
Back to browse
GitHub Repository
Minimalist, deterministic, lightweight, fast, and portable verification of AI-generated text: measure groundedness in a source as verbatim reuse (contiguous matching) or verbatim paraphrasing (LCS). Use as a decorator over generate(), a CLI or a CI gate. Useful for research and production purposes.
2 starsPython
I built a deterministic check for fabricated quotes in LLM output
by pobonin·Jul 14, 2026·3 points·0 comments
AI Analysis
●●●BangerBig BrainSolve My Problem
Deterministic verbatim checking beats probabilistic evals for catching hallucinated quotes.
Strengths
- •Decorator pattern enables real-time streaming verification without blocking generation loops.
- •Longest Common Subsequence algorithm detects paraphrasing where exact string matching fails.
- •Zero-dependency Python implementation ensures portability across diverse deployment environments.
Weaknesses
- •Strict verbatim logic may penalize valid semantic summarization in conversational agents.
- •No built-in visualization dashboard for analyzing grounding trends across large datasets.
Category
Target Audience
LLM application developers, RAG engineers
Similar To
Ragas · Arize Phoenix · LangSmith
Similar Projects
AI/ML●●Solid
ÆTHERYA Core – deterministic action-governance kernel for LLM agents
Fail-closed policy layer blocks LLM tool calls before execution, no LLM in decision path.
Big BrainNiche Gem
RobertMihai
103mo ago
AI/ML●●Solid
VigIA – A deterministic FSM in .NET 10 to stop LLM hallucinations
Deterministic FSM guards LLM outputs to prevent hallucination spirals in .NET.
Big BrainBold Bet
JordanCT
303mo ago
Developer Tools●●Solid
A deterministic middleware to compress LLM prompts by 50-80%
Deterministic prompt compression cuts tokens 50-80% without extra model calls.
Big BrainNiche Gem
rosspeili
303mo ago
AI/ML●●●Banger
A new benchmark for testing LLMs for deterministic outputs
Finally separates JSON validity from actual value hallucination in LLM outputs.
Big BrainSolve My Problem
khurdula
60302mo ago
Developer Tools●●Solid
Fullbleed – Rust HTML/CSS-to-PDF with Deterministic Output+Python CLI
It skips headless Chromium entirely and implements an HTML/CSS-to-PDF pipeline in Rust, exposing a Python wheel and CLI that releases the GIL and uses Rayon for parallel batch renders. The deterministic bits — fixed-point base unit, --repro-record/--repro-check, SHA256 outputs and vendored assets — are a clear, practical play for audited VDP/transactional workflows; what's still unknown is CSS spec coverage and whether subtle print-layout quirks will require hand-holding.
WizardryNiche Gem
krflol
204mo ago