This paper surveys scalable automated alignment methods for LLMs.
problem Scalability issues in traditional human-annotated alignment methods for LLMs.
method Categorizes and discusses various automated alignment methods.
result Emerging automated alignment methods are effective and scalable.
Hölder-DPO aligns models robustly with noisy human feedback.
problem No existing alignment methods can handle severe label noise.
method Proposes Hölder-DPO, a principled alignment loss with provable redescending property.
result Hölder-DPO enables scalable human feedback valuation and improves model alignment.
AI models aligned with human vision perform well on few data tasks.
problem Few-shot learning performance with limited data.
method Information-theoretic analysis and empirical testing of 491 models.
result Highly aligned models show better robustness to attacks and domain shifts.
Direct Density Ratio Optimization aligns LLMs with human preferences without assuming specific models.
problem Statistical inconsistency in aligning LLMs with human preferences.
method Direct Density Ratio Optimization (DDRO) estimates density ratio directly.
result DDRO is statistically consistent, converging to true human preferences as data grows.
PILAF optimizes reward models from human feedback for better policy alignment.
problem Creating accurate reward models from human feedback for policy optimization.
method Policy-Interpolated Learning for Aligned Feedback (PILAF) that explicitly aligns preference learning with maximizing underlying oracle reward.
result PILAF is optimal from both optimization and statistical perspectives, demonstrating strong performance in RLHF settings.
Paper explores limits and possibilities of aligning LLMs with human preferences.
problem Aligning LLMs with diverse human preferences to ensure fairness and informed outcomes.
method Analysis of probabilistic representation of human preferences and preservation of diverse preferences.
result LLMs can't fully align with human preferences using reward-based approaches due to Condorcet cycles, but mixed strategies are statistically possible.
The paper explores game-theoretic alignment of LLMs with human preferences, finding limitations and conditions.
problem Aligning LLMs with human preferences using game theory.
method Systematic study of payoff choices in a two-player zero-sum game for desirable alignment properties.
result Impossibility of preference matching in game-theoretic LLM alignment under standard assumptions.
Study compares DSPy teleprompter algorithms for aligning LLM evaluations with human annotations.
problem Aligning LLM evaluation metrics with human annotations.
method Comparative analysis of five teleprompter algorithms within the DSPy framework.
result Certain teleprompters outperform others in detecting hallucinations.
Develops methods to correct bias in AI feedback for more accurate alignment.
problem Systematic bias in AI feedback compared to human labels.
method Two debiased alignment methods: DDPO and DIPO.
result Methods improve alignment efficiency and performance close to human-labeled data.
Efficiently identifies good policies by choosing contexts for human feedback.
problem Efficiently acquiring human feedback for preference alignment in large language models.
method Formalizes active exploration as a dueling bandit problem and proposes an active exploration algorithm with a polynomial worst-case regret bound.
result Proposed method outperforms baselines with limited human preferences on various language models and datasets.
Geometric framework detects concept frustration between human concepts and machine representations.
problem Aligning human concepts with machine learning representations.
method Geometric framework and similarity measures for detecting concept frustration.
result Concept frustration affects machine learning model performance and reorganizes learned concept representations.
Proposes a robust algorithm for aligning large language models with human preferences.
problem Misspecification in preference models, reference policies, and reward functions.
method Doubly robust preference optimization algorithm.
result Superior and more robust performance compared to state-of-the-art algorithms.
Stable and consistent model alignment for language models without assuming human preference models.
problem Lack of statistical consistency in existing alignment methods.
method Relative density ratio optimization between preferred and mixture of preferred and non-preferred data distributions.
result Our approach achieves statistical consistency and stability, providing tighter convergence guarantees.
RAFT fine-tunes models using high-quality samples to align them with human preferences.
problem Aligning generative models with human ethics and preferences.
method RAFT selects high-quality samples, discards undesired behavior, and fine-tunes the model on filtered samples.
result RAFT improves model performance in reward learning and automated metrics.
SAIL improves online alignment of large language models with minimal feedback.
problem Offline RLHF methods often lead to sub-optimal performance due to fixed preference datasets.
method SAIL uses bilevel optimization and a single-level first-order method to iteratively refine model alignment.
result SAIL significantly improves alignment performance on open-sourced datasets with minimal computational overhead.
Enhances AI models with human feedback for noisy data.
problem Improving AI model alignment with human feedback in noisy environments.
method Two-stage SL+LHF framework connecting machine learning with human feedback.
result The LNCA ratio identifies conditions for SL+LHF superiority over pure SL.
A new method reduces the computational burden of safety alignment for large language models.
problem Safety concerns in large language models and the need to align them with human preferences.
method Optimal dualization approach to reduce constrained alignment to an unconstrained problem.
result Our algorithms MoCAN and PeCAN significantly reduce computational burden and improve training stability.
EBRM improves robustness and generalization of language model rewards.
problem Challenges in capturing complex human preferences and generalizing to unseen data in reward models.
method Energy-Based Reward Model (EBRM) that models reward distribution explicitly, using conflict-aware data filtering, label-noise-aware contrastive training, and hybrid initialization.
result Significant improvements in robustness and generalization, up to 5.97% improvement in safety-critical alignment tasks.
Framework ensures alignment between humans and machines in LLMs.
problem Human-machine misalignment in LLMs scoring mechanisms.
method Lightweight calibration framework for blackbox models.
result Provably guarantees alignment between humans and machines.
New RLHF approach mitigates bias in aligning LLMs with human preferences.
problem Algorithmic bias in RLHF leading to preference collapse.
method Preference Matching (PM) RLHF, using PM regularizer and conditional variant.
result 29% to 41% improvement in alignment with human preferences.
Paper shows equivalence between two alignment methods and introduces a new algorithm.
problem Ensuring human alignment of large language models for useful, safe, and pleasant user experience.
method Introduces IPO-MD algorithm, showing equivalence between IPO and Nash-MD methods.
result Equivalence between IPO and Nash-MD methods proven when considering online version of IPO.
High-dimensional neural network manifolds misalign with human perception, causing adversarial examples.
problem Adversarial attacks fool neural networks, but their origin is unclear.
method Defined and analyzed a network's perceptual manifold (PM) for a class concept.
result Neural network PMs have orders of magnitude higher dimensions than natural human concepts, suggesting exponential misalignment.
Conformal Alignment ensures trustworthy outputs from foundation models.
problem Ensuring outputs from foundation models align with human values in high-stakes tasks.
method A framework that trains an alignment predictor using reference data to select trustworthy outputs.
result Conformal Alignment accurately identifies trustworthy outputs via lightweight training over moderate reference data.
Unified framework for aligning LLMs from human feedback.
problem Lack of strong theoretical justification for RLHF and inability to compare methods.
method Reframed alignment as distribution learning from pairwise preferences, proposing three principled objectives.
result Proposed objectives achieve strong non-asymptotic convergence to target LM.
A Markov Chain approach for aligning generative models from pairwise human preferences.
problem Aligning generative models from pairwise human preferences.
method Markov Chain from Human Feedback (MCHF)
result MCHF converges geometrically fast to the stationary distribution.
Survey of alignment techniques for large language models.
problem Ensuring large language models align with human values.
method Analysis of diverse alignment methods and training paradigms.
result Preference-based methods offer more flexibility for nuanced alignment.
Dual active learning improves RLHF by selecting optimal conversations and teachers.
problem Efficiently aligning LLMs with human preferences using RLHF from feedback.
method Offline RL for conversation and teacher selection, dual active reward learning, pessimistic RL.
result The proposed algorithm achieves minimal generalized variance and outperforms state-of-the-arts.
Method tackles uncertainty in reward models for LLMs from heterogeneous human feedback.
problem Uncertainty in reward models for LLMs from heterogeneous human feedback.
method Heterogeneous preference framework and alternating gradient descent algorithm.
result Established theoretical guarantees for estimator convergence and asymptotic distribution.
Unified framework simplifies DPO algorithms for LLM alignment.
problem Vast number of DPO variants complicates model alignment.
method Mutual information inspired unifying framework with flexible priors.
result Many DPO variants can be derived from the new framework.
This paper introduces f-DPO, a generalized approach to Direct Preference Optimization using diverse divergence constraints.
problem Aligning large language models with human preferences while mitigating safety risks.
method Incorporates diverse divergence constraints to simplify the relationship between reward and optimal policy, eliminating the need for estimating the normalizing constant.
result Optimizes LLMs to align with human preferences more efficiently and under a broader set of divergence constraints.
New framework formalizes RLHF trilemma: improving safety, fairness, and robustness is computationally infeasible.
problem Aligning large language models with diverse human values while maintaining computational feasibility and robustness.
method Complexity-theoretic analysis integrating statistical learning theory and robust optimization.
result Achieving both representativeness (epsilon <= 0.01) and robustness (delta <= 0.001) for global-scale populations requires super-polynomial operations.
DRO-REBEL improves LLM alignment by robustly updating models online.
problem Overfitting and drifting of LLMs during RLHF.
method DRO-REBEL uses type-p Wasserstein, KL, and χ2 ambiguity sets for robust online updates. result DRO-REBEL achieves faster convergence and better performance than prior methods.
Deep models have advanced prediction in many domains, but their lack of interpretability remains a key barrier to the adoption in many real world applications. There exists a large body of work aiming to help humans understand these black box functions to varying levels of granularity -- for example, through distillati…
RLHF uses human feedback to train AI models, posing statistical challenges.
problem Aligning AI models with human preferences using noisy, subjective feedback.
method Supervised fine-tuning, reward modeling, policy optimization, statistical ideas.
result Statistical methods for reward function learning and policy optimization.
LLM evaluation suffers from systematic biases and lacks reliable positive judgments.
problem LLM evaluation suffers from systematic biases and lacks reliable positive judgments.
method Formulate LLM evaluation as a positive-unlabelled learning problem and propose a geometric auditing framework based on Partial Optimal Transport.
result Improved alignment with human preferences, increased robustness to presentation biases, and interpretable confidence estimates.
Framework enhances AI explainability by aligning with human cognitive models.
problem Lack of explainability in AI models hinders trust and accountability.
method Integrates explainability techniques with Malle's five category model of behavior explanation.
result Demonstrates practical relevance in credit risk assessment and regulatory analysis.
Improved text-to-image alignment using iterative VQA feedback.
problem Misalignment between text prompts and generated images, especially for complex inputs.
method Decompose complex prompts into assertions, evaluate each using VQA, combine scores iteratively.
result Significantly higher correlation with human ratings compared to CLIP, BLIP scores.
SPPO optimizes language model alignment by treating preferences as a game and achieving state-of-the-art performance.
problem Capturing intransitivity and irrationality in human preferences for accurate language model alignment.
method Self-play-based approach to identify Nash equilibrium policy through iterative policy updates.
result SPPO achieves state-of-the-art win-rate of 28.53% on AlpacaEval 2.0 without external supervision.
Inference-aware meta-alignment of LLMs reduces computational cost.
problem Aligning LLMs to diverse human preferences is challenging due to conflicting criteria.
method IAMA trains a base model to be aligned to multiple tasks via different inference-time alignment algorithms, using non-linear GRPO for optimization.
result IAMA enables effective alignment of LLMs to multiple criteria with limited computational budget.
The abstract explores connections between reinforcement learning, scaling, and diffusion.
problem Aligning reinforcement learning with human feedback and scaling techniques.
method Clarifying connections between reinforcement learning, scaling, and diffusion.
result Introducing a resampling approach for alignment and reward-directed diffusion models.
New framework transfers latent knowledge from weak to strong models.
problem Aligning superhuman LLMs with human feedback.
method Transfer learning framework using refinement approach.
result Proves weak-to-strong generalization is possible.
The paper proposes a method to align AI models using conformal risk control.
problem Aligning AI models to meet end-user requirements in non-generative settings.
method Post-processing a pre-trained model to better align with a subset of functions using conformal risk control.
result A probabilistic guarantee that the resulting conformal interval around a model contains a function approximately satisfying a desired property.
Develops a statistical framework to measure uncertainty in model rankings based on human preferences.
problem Uncertainty in model rankings based on human preferences due to mismatch between human and model preferences.
method Statistical framework using pairwise comparisons by humans and models to provide rank-sets for each model.
result Rank-sets constructed using only pairwise comparisons by strong models often do not cover the true ranking of human preferences.
Paper proposes efficient RLHF methods for LLMs using active queries.
problem Efficiently aligning LLMs with human preferences using RLHF.
method Formalizes RLHF as a dueling bandit problem, introduces APPO and ADPO algorithms.
result ADPO achieves similar performance to state-of-the-art methods with fewer queries.
New research shows label refinement and weak training have limitations for aligning LLMs.
problem Limitations of refinement methods for aligning large language models.
method Analyzed probabilistic assumptions and alternative approaches to label refinement and weak training.
result Label refinement and weak training suffer from irreducible error, leaving a performance gap.
Study reveals LLM personality patterns but lacks behavioral consistency.
problem Understanding and validating personality traits in LLMs.
method Characterized LLM personality across three dimensions: training dynamics, self-report validity, and intervention effects.
result Self-reported traits do not reliably predict behavior, and instructional alignment affects trait expression but not behavior.
Our goal is to design architectures that retain the groundbreaking performance of CNNs for landmark localization and at the same time are lightweight, compact and suitable for applications with limited computational resources. To this end, we make the following contributions: (a) we are the first to study the effect of…
Mitigates biases in reward models using variational inference.
problem Spurious correlations in reward models that align large language models with human preferences.
method Formulates data-generating process, identifies non-spurious latent variables, and uses variational inference to recover them.
result Effective mitigation of spurious correlation issues, yielding more robust reward models.