REVIEW 9 cited by
Some things are more CRINGE than others: Iterative Preference Optimization with the Pairwise Cringe Loss
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
Some things are more CRINGE than others: Iterative Preference Optimization with the Pairwise Cringe Loss
read the original abstract
Practitioners commonly align large language models using pairwise preferences, i.e., given labels of the type response A is preferred to response B for a given input. Perhaps less commonly, methods have also been developed for binary feedback, i.e. training models given labels of type response A is good or bad. We show how an existing performant binary feedback method, the Cringe Loss (Adolphs et al., 2022), can be generalized to the pairwise preference setting using a simple soft margin extension. Pairwise Cringe Loss is straightforward to implement and efficient to train, and we find it outperforms state-of-the-art preference optimization algorithms such as PPO and DPO on the AlpacaFarm benchmark. We show that iterations of training of our model are important for improved results, and that we can generalize DPO to Iterative DPO in the same way.
Forward citations
Cited by 9 Pith papers
-
Self-Rewarding Language Models
Iterative self-rewarding via LLM-as-Judge in DPO training on Llama 2 70B improves instruction following and self-evaluation, outperforming GPT-4 on AlpacaEval 2.0.
-
Multi-Turn On-Policy Distillation with Prefix Replay
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.
-
World Feedback for Clinical Agents: Diagnosing RL in FHIR Environments
MedAgentBench-v3 shows capability ceilings and format-knowledge barriers limit pure RL to 18.2% while rule-based SFT reaches 34.1% on clinical protocol tasks.
-
Self-Improvement Can Self-Regress: The Rise-and-Collapse Failure Mode of LLM Self-Training
REINFORCE self-training on competitive programming tasks exhibits robust rise-then-collapse in pass@1; CARE, ES, and GRPO mitigate it in model-size-dependent ways across Qwen-2.5-3B/7B and a Gemma pilot.
-
UNA: A Unified Supervised Framework for Efficient LLM Alignment Across Feedback Types
UNA unifies binary, pairwise, and score-based feedback for LLM alignment via a generalized implicit reward function shown optimal by the log sum inequality.
-
Multi-Turn On-Policy Distillation with Prefix Replay
By replaying teacher prefixes with a step-decay schedule, multi-turn on-policy distillation can run without live environment interaction, matching or slightly beating online OPD accuracy.
-
Boosting Reinforcement Learning with Verifiable Rewards via Randomly Selected Few-Shot Guidance
FEST improves RLVR sample efficiency on math and coding benchmarks by combining supervised signals, on-policy signals, and decaying weights on just 128 randomly chosen demonstrations, matching full-dataset baselines.
-
Failure Modes of Maximum Entropy RLHF
Derives SimPO from MaxEnt RL and reports that MaxEnt RL in online RLHF exhibits frequent overoptimization and unstable KL dynamics across scales, unlike stable KL-constrained baselines.
-
PoliLegalLM: A Technical Report on a Large Language Model for Political and Legal Affairs
PoliLegalLM, trained with continued pretraining, progressive SFT, and preference RL on a legal corpus, outperforms similar-scale models on LawBench, LexEval, and a real-world PoliLegal dataset while staying competitiv...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.