Pith. sign in

REVIEW 8 cited by

Beyond Reverse KL: Generalizing Direct Preference Optimization with Diverse Divergence Constraints

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2309.16240 v1 pith:ZAFFDLPL submitted 2023-09-28 cs.LG cs.AIstat.ML

classification cs.LGcs.AIstat.ML
keywords divergenceconstraintsdivergencesalignmentrewardunderapproachdirect
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

The increasing capabilities of large language models (LLMs) raise opportunities for artificial general intelligence but concurrently amplify safety concerns, such as potential misuse of AI systems, necessitating effective AI alignment. Reinforcement Learning from Human Feedback (RLHF) has emerged as a promising pathway towards AI alignment but brings forth challenges due to its complexity and dependence on a separate reward model. Direct Preference Optimization (DPO) has been proposed as an alternative, and it remains equivalent to RLHF under the reverse KL regularization constraint. This paper presents $f$-DPO, a generalized approach to DPO by incorporating diverse divergence constraints. We show that under certain $f$-divergences, including Jensen-Shannon divergence, forward KL divergences and $\alpha$-divergences, the complex relationship between the reward and optimal policy can also be simplified by addressing the Karush-Kuhn-Tucker conditions. This eliminates the need for estimating the normalizing constant in the Bradley-Terry model and enables a tractable mapping between the reward function and the optimal policy. Our approach optimizes LLMs to align with human preferences in a more efficient and supervised manner under a broad set of divergence constraints. Empirically, adopting these divergences ensures a balance between alignment performance and generation diversity. Importantly, $f$-DPO outperforms PPO-based methods in divergence efficiency, and divergence constraints directly influence expected calibration error (ECE).

Discussion (0). Sign in to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Semiparametric Preference Optimization: Your Language Model is Secretly a Single-Index Model

    cs.LG 2025-12 conditional novelty 7.0 of 10

    Preference-based LLM alignment under an unknown reward-preference link becomes a single-index model; three new algorithms converge to the optimal divergence-constrained policy without knowing the link.

  2. Multi-Turn On-Policy Distillation with Prefix Replay

    cs.LG 2026-07 conditional novelty 6.0 of 10

    ReOPD offline-distills multi-turn agentic LLMs via teacher-prefix replay plus step-decay sampling, matching online OPD accuracy at ≥4× speed with zero tool calls.

  3. Distribution Preference Optimization: A Fine-grained Perspective for LLM Unlearning

    cs.LG 2025-10 conditional novelty 6.0 of 10

    DiPO is a distribution-level unlearning method that constructs preference distributions from the model's own high-confidence logits and achieves state-of-the-art forget quality on TOFU while preserving utility.

  4. Thompson Sampling in Online RLHF with General Function Approximation

    cs.LG 2025-05 reject novelty 6.0 of 10

    A model-free posterior sampling algorithm for online RLHF is shown to achieve O(sqrt(T)) regret when the completed function class has low Bellman eluder dimension.

  5. Square$\chi$PO: Differentially Private and Robust $\chi^2$-Preference Optimization in Offline Direct Alignment

    cs.LG 2025-05 conditional novelty 6.0 of 10

    SquareχPO, a square-loss variant of χPO, achieves optimal 1/sqrt(n) suboptimality under label privacy and Huber corruption for offline direct alignment with general function classes.

  6. Normalized Rewards for Preference Optimization

    cs.LG 2026-06 conditional novelty 5.0 of 10

    A regularization term that conserves the combined length-normalized probability of chosen and rejected responses reduces likelihood displacement in DPO/SimPO, improves AlpacaEval and benchmark outcomes, and acts prima...

  7. Token-level Accept or Reject: A Micro Alignment Approach for Large Language Models

    cs.CL 2025-05 reject novelty 5.0 of 10

    MARA aligns LLMs with human preferences by training a 4M-parameter MLP to accept or reject candidate tokens, avoiding full-model fine-tuning, with measured gains based on the same reward models used in training.

  8. The Hitchhiker's Guide to Agentic AI: From Foundations to Systems

    cs.AI 2026-06 unverdicted novelty 2.0 of 10

    A survey-style reference book mapping the full agentic-AI stack from transformer internals to production deployment, with no new research result.

Pith tools