English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

OptimismBench: Directional Bias in LLM Judgment — Why 70% + 15% Doesn't Equal 100%

Forum topic · ✨步子哥 · 2026-08-03

Summary

A Zhichai.net forum post reviews the paper OptimismBench: Forecasting Bias and the Alignment Effect in Language Model Judgment (arXiv:2607.26981, Cho Seonglae & Koshiyama Adriano). The paper introduces a ground-truth-free method for detecting directional bias in LLM probabilistic judgment: 'inverted pairs' ask P(success) and P(failure) for the same scenario, and their deviation from 100% (the Skew metric) measures optimism or pessimism. Across 16 models from 8 vendors, 14 showed optimism bias, while only two Anthropic frontier (Claude) models were pessimistic. Comparing 11 base-vs-chat model pairs revealed that post-training determines bias direction: Qwen's fine-tuning shifted models toward optimism, Llama's toward pessimism. A cross-language experiment (17 models, 6 languages) found model identity matters 4.7x more than language, and bias can flip sign in recommendation tasks, implying bias is a model-by-task-framing property. The post argues calibration metrics (ECE, Brier) are structurally blind to signed errors, that alignment training tilts probabilities as a side effect, and that downstream pipelines inheriting this tilt should audit directional bias in risk assessment, forecasting, and planning use cases.

This post is a GEO-optimized version of the original topic on zhichai.net.

Paper: OptimismBench: Forecasting Bias and the Alignment Effect in Language Model Judgment Authors: Cho Seonglae, Koshiyama Adriano arXiv: 2607.26981

Key points

  • The puzzle: Ask an LLM for P(startup succeeds) → 70%. Ask P(it fails) → 15%. 70 + 15 = 85. The missing 15 points are not rounding error — they reveal a directional bias that calibration metrics (ECE, Brier score) cannot catch, because those metrics cancel out signed errors.
  • Ground-truth-free bias detection: OptimismBench's 'inverted pair' method compares the model's answers to two phrasings of the same scenario (P(success) vs P(failure)). If the model is unbiased, they sum to 100%. The deviation defines the Skew metric: Skew > 0 = optimism, Skew < 0 = pessimism. No ground truth required.
  • Four tracks: calibration control (15 questions with real frequencies), probability estimation (main inverted-pair track), recommendation (judgment replaced by recommending), and salience (varying scenario salience).
  • Interventions fail: narrative manipulation, perspective shifting, anchoring gradients, and self-debiasing were all tested — bias survived every intervention.
  • Main findings

  • 14 of 16 models (8 vendors) are optimistic; only 2 are pessimistic — both from Anthropic's frontier Claude line.
  • Post-training determines bias sign: across 11 base-vs-chat pairs in 4 families, Qwen's post-training shifted models from pessimistic to optimistic; Llama's shifted from optimistic to pessimistic. Bias is a product of alignment training, not architecture or pretraining.
  • Model identity > language: across 17 models and 6 languages, between-model variance was 4.7x between-language variance.
  • Bias flips by task framing: in the recommendation track, a model optimistic in probability judgment could become pessimistic. Bias is a joint property of model × task framing, not a fixed trait.
  • Why it matters

    1. Structural blind spot of calibration metrics: ECE and Brier score are unsigned — a model can be well-calibrated on average while systematically overestimating positive outcomes. Directional bias hides in the *direction* of error. 2. Alignment is not neutral: RLHF/DPO shapes not just safety and helpfulness but probability tilt. A model trained to be 'more helpful' may systematically overestimate success rates — dangerous for risk assessment, medical decisions, and financial forecasting. 3. Downstream inheritance: the paper's closing line:

    > "When alignment makes a model more helpful, it also tilts its probabilities; downstream pipelines inherit the tilt by default."

    Any system using LLMs for probabilistic judgment should audit the model's directional bias rather than assume 'the model said 70%, so it's 70%.'

    Cross-domain analogy: human optimism bias

    The title's 'Optimism' echoes the classic human optimism bias (Weinstein, 1980): people systematically overestimate positive events. LLMs appear to have absorbed this human cognitive pattern — but with a key difference: human bias is universal, while LLM bias direction is set by post-training, meaning alignment could counter-calibrate it. This raises an open question: should alignment target 'unbiased' or 'human-like optimism'? For competing with humans in prediction markets, the latter may be more accurate; for risk assessment, unbiased is correct.

    Limitations (acknowledged by the authors)

  • The family-direction conclusion rests mainly on the Qwen and Llama pairs; Gemma-2-2b and Mistral-Small-24B are single-pair pilots with limited statistical power.
  • Cross-language evidence is unbalanced: 6 native-language prompts plus 4 English system prompts (DE/FR/HI/JA), a known confounder.
  • Skew measures internal consistency, not agreement with human judgment — an 'unbiased' model is not necessarily an 'accurate' one.
  • No matched human-subject baseline, despite citing Weinstein (1980).

Commenter's take

The methodological elegance is reframing: instead of comparing predictions to reality, compare the model to *itself*, sidestepping the ground-truth dependency. The practical worry: we increasingly use LLMs as probability-judgment engines, but their bias carries 'family traits' — a risk system built on Claude vs. one built on Qwen may point in opposite directions. And there is a real tension between alignment and calibration: helpfulness rewards optimistic answers, accuracy punishes them. The tension may have no perfect solution, but at least it can now be measured.

Links: Paper · HTML version · Dataset: 3,870 test items, 10 languages, open-sourced.

Tags

#llm-evaluation#optimism-bias#calibration#alignment#optimismbench#directional-bias#ai-safety#paper-review

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178503894