AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Global Merger-Arbitrage Forecasting with Language Models

Hinal Jajal, Michał Mucha, Charles Sweat, Chris Pulman, Charlie Flanagan, Peter Anderson

ICML 2026 · Large language models › Language, safety and interpretability · approach: Faithfulness of LLM explanations · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
finetune GPT-4o with hindsight supervisionGPT-4o · Merger outcome +/- predictionclass-balanced Brier: 0.201 -> 0.151Improved
add isotonic post-hoc calibration instead of finetuningGPT-4o · Merger outcome +/- predictionclass-balanced Brier: 0.201 -> 0.186Improved
add full training data (vs 50%) (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.177 -> 0.151Improved
add oversampling of terminated deals (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.170 -> 0.151Improved
oversample more aggressively (2.0, 1.65) (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.151 -> 0.163Hurt
add hindsight-guided postmortem supervision (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.183 -> 0.151Improved
add market-probability smoothing (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.181 -> 0.151Improved
add specialized research agents to Deal Card (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.172 -> 0.151Improved
add Stakeholder Ownership agent to Deal Card only (ablation)GPT-4o-FT · Merger outcome +/- predictionclass-balanced Brier: 0.172 -> 0.158Improved
finetune GPT-4oGPT-4o · Days to completionMAPE: 0.463 -> 0.385Improved
finetune GPT-4o with hindsight supervisionGPT-4o · Rubric coverage of Day-1 reportsrubric coverage: 21.5% -> 50.2%Improved
finetune GPT-4oGPT-4o · Deal report groundingunsupported-claim rate: 0.3% -> 0.1%No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy