logprobs

An experimental approach for AI search visibility.

Interactive Streamlit app that visualises token-level log probabilities from Azure OpenAI

Token-level entropy trajectories from LLM logprobs. Models can measure their own uncertainty — grounding it in truth requires external verification.

A drop-in confidence gate for LLM agents: turn token logprobs into an accept / verify / ask-a-human routing decision, with honest per-model calibration. Local & open-model first, plus an MCP server.

Single-pass CES entropy risk scoring for LLM outputs with calibration, provider logprobs, CLI, and agent/RAG guardrails.

A demo to show confidence probabilities in AI responses

An interactive client for exploring token-level log probabilities from OpenAI-compatible chat models.

Minimal demo of calibrated LLM-as-a-judge scores from the single-call token-logprob distribution (Wang et al. 2025; G-Eval).

Renders every token a model generated with the tokens it rejected, ghosted above it by probability

Stream LLM tokens with logprobs, compute per-token Shannon entropy, render a color-coded HTML uncertainty map

MSc thesis: mechanistic interpretability of Chain-of-Thought reasoning in LLMs via SAE features and logprob signals

Measuring the computational cost of deception in LLMs through logprob analysis

Analyze LLM token log-probabilities: perplexity, per-token confidence, low-confidence span detection, and a terminal heatmap. Zero deps, CLI + library.

Folded into cardinal-harness (src/seriate). 0.1.2 stays published; name parked. Provenanced single-token LLM elicitation instruments.

LLM token visualizer for decoder risk, logprobs, embeddings, branching, and consistency checks.