RANK ANALYSIS OF INCOMPLETE BLOCK DESIGNS THE METHOD OF PAIRED COMPARISONS

By R. Bradley, M. E. Terry. 1952

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

Citations in the corpus

Direct Preference Optimization: Your Language Model is Secretly a Reward Model (2023, methodology): “Like existing algorithms, DPO relies on a theoretical preference model (such as the Bradley-Terry model; [5]) that measures how well a given reward function aligns with empirical preference data.”
Improving alignment of dialogue agents via targeted human judgements (2022, methodology): “Preference reward models As described in section 2.5, our preference RMs are all Bradley-Terry (Elo) models (Bradley and Terry, 1952) of the same form as Ziegler et al. (2019).”
Generalized Results for the Existence and Consistency of the MLE in the Bradley-Terry-Luce Model (2021, methodology): “Among them, the renown Bradley-Terry-Luce (BTL) model (Bradley & Terry, 1952; Luce, 2012) has experienced a surge in popularity and use, due to its high interpretability and expressive power.”
From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline (2024, methodology): “Following Chatbot Arena, we adopt the Bradley-Terry [5] model to produce model’s the final model scores.”
Offline A/B Testing for Recommender Systems (2018, methodology): “Good examples of distributions are the BradleyTerry Luce (BTL) model [3], the Placket-Luce model [4, 7, 20] or more generally Thurstonian models [26, 28].”
Extrapolating Beyond Suboptimal Demonstrations via Inverse Reinforcement Learning from Observations (2019, methodology): “This form of loss function follows from the classic Bradley-Terry and Luce-Shephard models of preferences (Bradley & Terry, 1952; Luce, 2012) and has been shown to be effective for training neural networks from preferences (Christiano et al., 2017; Ibarz et al., 2018).”
Preference-based Reinforcement Learning with Finite-Time Guarantees (2020, unknown): “The comparisons are generated following the Bradley-Terry-Luce model [9]: φ(τ1, τ2) = 1 1+exp(−(r(τ1)−r(τ2))/c) , with c being either 0.”
Statistical ranking and combinatorial Hodge theory (2008, background): “The third and fourth models are known as linear model [32] and Bradley-Terry model [6] respectively in the statistics and psychology literature.”
RewardBench: Evaluating Reward Models for Language Modeling (2024, methodology): “…model involves training a classifier to predict the human preference probability, p ∗ , between two answers, as modeled by a Bradley-Terry model (Bradley and Terry, 1952): Then, estimate the parameters of the reward model by optimizing the maximum likelihood loss as follows: For language…”
Preference-based interactive multi-document summarisation (2019, methodology): “The Bradley-Terry (BT) model (Bradley and Terry 1952) is a widely used preference learning model, which approximates the ranking σ ∗ x by approximating the utility function U ∗ x : Suppose we have observed N preferences: , where y i, 1 , y i, 2 ∈ Y x are the summaries presented to the oracle in the…”