Pith. sign in

REVIEW 8 cited by

Advancing LLM Reasoning Generalists with Preference Trees

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 2404.02078 v1 pith:CXF2UOE4 submitted 2024-04-02 cs.AI cs.CLcs.LG

classification cs.AIcs.CLcs.LG
keywords reasoningpreferencelearningmodelstasksultrainteracteurusbenchmarks
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We introduce Eurus, a suite of large language models (LLMs) optimized for reasoning. Finetuned from Mistral-7B and CodeLlama-70B, Eurus models achieve state-of-the-art results among open-source models on a diverse set of benchmarks covering mathematics, code generation, and logical reasoning problems. Notably, Eurus-70B beats GPT-3.5 Turbo in reasoning through a comprehensive benchmarking across 12 tests covering five tasks, and achieves a 33.3% pass@1 accuracy on LeetCode and 32.6% on TheoremQA, two challenging benchmarks, substantially outperforming existing open-source models by margins more than 13.3%. The strong performance of Eurus can be primarily attributed to UltraInteract, our newly-curated large-scale, high-quality alignment dataset specifically designed for complex reasoning tasks. UltraInteract can be used in both supervised fine-tuning and preference learning. For each instruction, it includes a preference tree consisting of (1) reasoning chains with diverse planning strategies in a unified format, (2) multi-turn interaction trajectories with the environment and the critique, and (3) pairwise data to facilitate preference learning. UltraInteract allows us to conduct an in-depth exploration of preference learning for reasoning tasks. Our investigation reveals that some well-established preference learning algorithms may be less suitable for reasoning tasks compared to their effectiveness in general conversations. Inspired by this, we derive a novel reward modeling objective which, together with UltraInteract, leads to a strong reward model.

Discussion (0). Continue with ORCID 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. 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.

  2. SC2Arena and StarEvolve: Benchmark and Self-Improvement Framework for LLMs in Complex Decision-Making Tasks

    cs.LG 2025-08 conditional novelty 6.0 of 10

    The authors propose SC2Arena, a full-coverage StarCraft II benchmark for LLMs, and StarEvolve, a planner-executor-verifier self-improvement framework, claiming superior strategic planning.

  3. Measuring Diversity in Synthetic Datasets

    cs.CL 2025-02 conditional novelty 6.0 of 10

    DCScore measures dataset diversity as the sum of self-classification probabilities under a softmax similarity matrix, and the paper shows it tracks generation temperature, human judgment, and LLM rankings.

  4. MoGU V2: Toward a Higher Pareto Frontier Between Model Usability and Security

    cs.CL 2025-09 conditional novelty 5.0 of 10

    MoGUv2 embeds small routers in the deeper layers of LLMs to dynamically blend a helpful variant and a refusal variant, improving safety against jailbreak and fine-tuning attacks while preserving usability.

  5. Anchoring Refusal Direction: Mitigating Safety Risks in Tuning via Projection Constraint

    cs.CL 2025-09 conditional novelty 5.0 of 10

    ProCon anchors each sample's hidden-state projection onto the LLM's initial refusal direction during instruction fine-tuning, reducing refusal-direction drift and safety risks with limited task-performance loss.

  6. On the Effect of Negative Gradient in Group Relative Deep Reinforcement Optimization

    cs.LG 2025-05 conditional novelty 5.0 of 10

    GRPO can suffer from Lazy Likelihood Displacement, where correct responses gain little or lose likelihood, and selectively downweighting penalties on harmful tokens (NTHR) mitigates this and boosts math reasoning performance.

  7. Towards Reliable, Uncertainty-Aware Alignment

    cs.LG 2025-07 reject novelty 4.0 of 10

    Variance-aware RLHF adds a variance-weighted KL penalty to PPO and reduces reward variance and the risk of underperforming the reference policy in the paper's experiments.

  8. DPO-Shift: Shifting the Distribution of Direct Preference Optimization

    cs.CL 2025-02 conditional novelty 4.0 of 10

    DPO-Shift multiplies the rejected-response term in the DPO loss by a factor f(λ) below 1, shifting chosen-response probabilities upward at the expense of some reward margin.

Pith tools