Prompt EngineeringTamera Lanham et al. (Anthropic)Anthropic

Measuring Faithfulness in Chain-of-Thought Reasoning published

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Measuring Faithfulness in Chain-of-Thought Reasoning published — applies interventions such as inserted errors and truncated sentences to the CoT and measures how the answer changes

Why it matters

A methodology that quantified faithfulness through intervention experiments

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2307.13702

Added in Week 35, 2026 · See that week

On the map — ① Prompt Engineering line

  1. Previous stopPaLM 2
  2. This recordMeasuring Faithfulness in Chain-of-Thought Reasoning published
  3. Next stopChain-of-Verification
See it on the map

Same week

Meta releases Llama 2 — commercial use allowed, but under a Community LicensePressFlashAttention-2 publishedPaper

Nearby on the same axis

Anthropic releases Claude 2 and opens claude.ai in betaOfficialSkeleton-of-Thought publishedPaperAnthropic releases Claude Instant 1.2OfficialGraph of Thoughts publishedPaper