Pith. sign in

REVIEW 10 cited by

DistiLLM-2: A Contrastive Approach Boosts the Distillation of LLMs

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 2503.07067 v2 pith:6P32H7LT submitted 2025-03-10 cs.CL cs.AIcs.LG

classification cs.CLcs.AIcs.LG
keywords modelsstudentapproachcontrastivedatadistillationdistillm-2across
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Despite the success of distillation in large language models (LLMs), most prior work applies identical loss functions to both teacher- and student-generated data. These strategies overlook the synergy between loss formulations and data types, leading to a suboptimal performance boost in student models. To address this, we propose DistiLLM-2, a contrastive approach that simultaneously increases the likelihood of teacher responses and decreases that of student responses by harnessing this synergy. Our extensive experiments show that DistiLLM-2 not only builds high-performing student models across a wide range of tasks, including instruction-following and code generation, but also supports diverse applications, such as preference alignment and vision-language extensions. These findings highlight the potential of a contrastive approach to enhance the efficacy of LLM distillation by effectively aligning teacher and student models across varied data types.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 10 Pith papers

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

  1. Poly-OPD: Heterogeneous Multi-Teacher On-Policy Distillation for Capability-Selectable Flow Models

    cs.CV 2026-08 conditional novelty 7.0 of 10

    Poly-OPD distills two heterogeneous text-to-image teachers into one 2.5B student via a pixel bridge and DINOv2-space supervision, surpassing both teachers on GenEval.

  2. Cross-Tokenizer On-Policy Distillation via Byte-Prefix Marginalization

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Byte-Prefix Marginalization maps a teacher's next-token distribution onto the student's vocabulary through shared byte prefixes plus an explicit residual, giving a mass-preserving target for on-policy distillation acr...

  3. When Top-K Misses the Decision: Tool-Call Drift in Multi-Teacher On-Policy Distillation

    cs.CL 2026-07 conditional novelty 7.0 of 10

    Top-K teacher logits can preserve almost all probability mass while dropping the low-probability '<tool call>' token, biasing distilled models to over-use tools; restoring the omitted token reduces over-calling 14.2%→...

  4. Weak-to-Strong Generalization via Direct On-Policy Distillation

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Transferring the log-ratio of a small model's pre-RL and post-RL checkpoints provides a dense implicit reward that improves stronger student models at a fraction of the cost of direct RL.

  5. A Unified Approach to Interpreting Knowledge Distillation for Large Language Models via Interactions

    cs.LG 2026-05 conditional novelty 6.5 of 10

    KD sparsifies complex interactions in student LLMs; a Complex Interaction Penalty that enforces this improves multiple KD methods on in- and out-of-domain benchmarks.

  6. Any-OPD: Heterogeneous On-Policy Distillation for Flow-Matching Models via Representation-Space Bridging

    cs.LG 2026-08 conditional novelty 6.0 of 10

    A distillation method that supervises a student on its own samples using teacher refinements compared in a frozen DINOv2 feature space, reporting preference-metric gains on a FLUX.1-dev to SD3.5-Medium transfer.

  7. Compete Then Collaborate: Frontier AI Teachers Build a Verifiable Curriculum to Improve a Coding Student Beyond Imitation

    cs.AI 2026-07 conditional novelty 6.0 of 10

    Imitation of multi-teacher verified coding solutions degrades a competent student, while RL with verifiable rewards on the same curriculum improves it.

  8. Focusing by Contrastive Attention: Enhancing VLMs' Visual Reasoning

    cs.CV 2025-09 conditional novelty 5.0 of 10

    CARVE contrasts attention maps from a general prompt and a specific question to mask out visual noise, then re-asks the question on the cropped and enlarged image, improving VQA accuracy by up to 75% on some benchmarks.

  9. Enhancing Reasoning Capabilities in SLMs with Reward Guided Dataset Distillation

    cs.AI 2025-06 conditional novelty 4.0 of 10

    AdvDistill uses group relative advantages computed from rule-based rewards to weight teacher responses during distillation, reportedly improving a 1.5B student on math tasks beyond its 7B teacher.

  10. Being Strong Progressively! Enhancing Knowledge Distillation of Large Language Models through a Curriculum Learning Framework

    cs.CL 2025-06 conditional novelty 4.0 of 10

    POCL wraps LLM knowledge distillation in a curriculum that increases data difficulty and temperature over stages, improving Rouge-L on small GPT-2 and OPT students, though ablations show temperature drives the gains.

Pith tools