Pith. sign in

REVIEW 4 major objections 5 minor 78 references

A 10-billion-parameter open-weight reasoning model claims the best accuracy-per-token balance in its class, producing roughly one-third as many reasoning tokens as Qwen3.5-9B and about 35% of OLMo-3-7B-Think's budget while staying competiti

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-02 14:03 UTC pith:6ZEPOX7A

load-bearing objection Solid system report with a genuinely useful token-efficiency analysis, but the headline numbers may describe a checkpoint other than the released weights — fix that and it earns its place. the 4 major comments →

arxiv 2607.20448 v1 pith:6ZEPOX7A submitted 2026-05-13 cs.CL cs.LG

Domyn-Small: A European 10B Reasoning Language Model

Simone Angarano , Francesco Bertolotti , Federico D'Ambrosio , Michele Resta , Alessandro Rognoni , Nicol\`o Ruggeri , Dario Salvati , Andrea Valenti
show 2 more authors
Alberto Veneri Martin Cimmino
This is my paper
classification cs.CL cs.LG
keywords reasoning language modeltoken efficiencyGRPODPOcontinued pre-traininginstruction followingopen weightsEU AI Act
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces Domyn-Small, a 10B open-weight reasoning model built by adapting a 2024-vintage foundation model through five post-training stages: continued pre-training, supervised fine-tuning with math annealing, GRPO with verifiable rewards, DPO, and multi-environment GRPO. Its central claim is that this pipeline achieves the strongest accuracy-efficiency balance in the 7–10B parameter class, measured as mean generated tokens per problem alongside accuracy. The model produces roughly one-third the reasoning tokens of Qwen3.5-9B and 35% of OLMo-3-7B-Think's budget, while posting competitive scores on IFEval (79.9) and GPQA-Diamond (50.0). If true, this matters because regulated enterprises seeking sovereign, cost-effective deployment could serve a capable reasoning model on a modest GPU allocation at substantially lower inference cost. The paper releases weights, a recipe-level post-training description, and an open-source inference framework.

Core claim

The paper's central claim is that a carefully sequenced post-training pipeline can convert an existing 10B base model into a reasoning model that sits on the accuracy/token-budget Pareto frontier of its class. The authors report that Domyn-Small generates a weighted grand mean of 2,690 tokens per problem on reasoning benchmarks—approximately 32% of Qwen3.5-9B's 8,440 and 35% of OLMo-3-7B-Think's 7,641—while reaching 93.2 on MATH-500, 35.7 on AIME 2025, and 50.0 on GPQA-Diamond. They attribute this efficiency to the SFT stage biasing toward shorter validated reasoning traces and to a soft length penalty in the GRPO reward. The final multi-environment GRPO stage, spanning math, code, QA, instr

What carries the argument

The central evaluation construct is the accuracy/token-budget Pareto frontier, with token efficiency treated as a first-class deployment metric: mean generated tokens per problem is reported alongside accuracy, and the two are read jointly as cost per correct answer. The mechanisms that move the model along this frontier are the GRPO soft length penalty (linearly decreasing from 0 to −0.25 over response length) and the SFT mixture's bias toward concise reasoning traces, supplemented by a multi-environment GRPO stage that uses leave-one-out advantage normalization and dynamic sampling to discard zero-variance groups.

Load-bearing premise

The headline benchmark numbers are measured on the latest checkpoint of the multi-environment GRPO stage, which the paper itself says may differ from the final released weights, so if the released v1.0 checkpoint produces different token counts or accuracy, the central efficiency claim describes a different model.

What would settle it

Reproduce the evaluation on the publicly released v1.0 weights using the paper's stated harness and decoding parameters, and compare mean tokens per problem on MATH-500, AIME 2025, and GPQA-Diamond. If the released model generates substantially more than the reported 2,690-token reasoning grand mean or scores far from 93.2, 35.7, and 50.0, the central claim collapses.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • At matched accuracy thresholds on reasoning tasks, Domyn-Small consumes roughly one-third the inference compute of Qwen3.5-9B and under 35% of OLMo-3-7B-Think's, reducing latency and leaving more headroom in fixed-context agentic loops.
  • The dual-mode reasoning toggle yields large within-model gains on code generation (+26.8 on HumanEval, +22.2 on MBPP) and science reasoning (+10.0 on GPQA-Diamond), confirming that explicit thinking traces matter most for multi-step search and synthesis.
  • The multi-environment GRPO stage recovers the instruction-following and tool-use regressions introduced by math-only GRPO, ending above the SFT baseline on IFEval and within one point on BFCL.
  • The 32K native context extends to 128K via YaRN, but 64K retrieval accuracy drops to 29.6 on RULER, indicating that training-free extrapolation underperforms native long-context training and pointing to a future long-context CPT phase.
  • The full post-training recipe and weights are released, allowing others to reproduce or adapt the pipeline on different base models and domains.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the token-efficiency advantage generalizes beyond the ten reported benchmarks, token-budget-aware post-training could become a standard objective for small-model deployment, shifting competition from raw accuracy toward cost-per-correct-answer.
  • The claimed balance depends on deployment granularity: Qwen3.5-9B still leads heavily on hard math (AIME 2025: 90.0 vs 35.7), so enterprises that need frontier math may not accept the trade for token savings.
  • A direct testable extension is to measure whether the efficiency advantage persists on multi-turn tool-calling trajectories, where the paper reports strong single-turn BFCL scores but a weak Multi-Turn split (7.0).
  • The 2024-vintage base model likely caps the ceiling; a modern reasoning-optimized base combined with the same pipeline could potentially preserve the token efficiency while closing the hard-math gap.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper introduces Domyn-Small, a 10B-parameter open-weight reasoning model obtained by adapting Italia 10B through a five-stage pipeline: continued pre-training on 503B tokens with context extension to 32K, SFT on a 3.85M-sample instruction mixture, math-only GRPO with verifiable rewards, DPO, and a final multi-environment GRPO across five task domains. The central claim is a strong accuracy-efficiency balance in the 7–10B class: the model is reported to produce roughly one-third the tokens of Qwen3.5-9B and about 35% of OLMo-3-7B-Think's token budget on core reasoning benchmarks, with competitive GPQA-Diamond (50.0) and IFEval (79.9). The paper also releases the post-training recipe, safety evaluation, EU AI Act documentation, and the Domyn Swarm inference framework. The headline efficiency/accuracy positioning is conditional on the evaluated checkpoint being the same as the released v1.0 weights, an issue the manuscript itself raises.

Significance. If the reported numbers are reproducible on the released weights, this is a useful contribution: a compact open-weight reasoning model with a fully documented post-training recipe, MIT-licensed weights, and an open-source HPC inference framework. The paper is unusually transparent about data mixtures, hyperparameters, and stage-wise evaluation, and it ships a concrete artifact. However, two load-bearing issues prevent immediate acceptance: the evaluation checkpoint may not be the released v1.0 checkpoint, and no contamination analysis is provided for benchmarks that overlap with the large public SFT/CPT mixtures. The efficiency claim, which is the paper's main novelty, is only as strong as the benchmark/token-count measurements, so these issues must be resolved in revision.

major comments (4)
  1. [Section 5.1 vs Section 4.4 and Section 6.1] The manuscript is internally inconsistent about which weights were evaluated. Section 5.1 states: "The Domyn-Small results presented here correspond to the latest checkpoint of the multi-environment GRPO stage (Section 4.4); values may be refreshed upon final model release." Section 4.4 says released v1.0 weights are "the best-validation checkpoint" from that run, which is not necessarily the latest checkpoint. Section 6.1, by contrast, says safety evaluations were run on "the released Domyn-Small-v1.0 checkpoint." If the latest checkpoint differs from the best-validation checkpoint, then Tables 7–9, Figure 3, and the abstract's efficiency numbers describe an unreleased model, not the downloadable artifact. This must be fixed by either re-evaluating the released v1.0 weights for all headline results or explicitly confirming that the latest checkpoint is identical to the released best-val
  2. [Section 4.1 / 5.1] No contamination analysis is reported for any evaluation benchmark. The SFT mixture is assembled from 40+ public instruction datasets (Section 4.1, Table 2), and the CPT corpus includes SFT-style instruction data, web crawl, and academic text (Figure 2). The evaluation suites include public benchmarks such as GPQA-Diamond, MMLU, MMLU-Pro, IFEval, HumanEval, and BFCL. Without an n-gram overlap or holdout analysis, the reported accuracy numbers—and therefore any accuracy-efficiency conclusion—may be inflated by training-set leakage. This is a missing load-bearing support, not a matter of style. Please report contamination checks for all headline benchmarks and, if needed, re-score on decontaminated subsets.
  3. [Section 5.1 / Table 9] The evaluation mixes sampling protocols in a way that undermines the precision of the efficiency claim. Section 5.1 says math benchmarks use avg@48 sampling, but Table 9's footnote says "thinking-on AIME 2025 is reported as avg@48, all other thinking-on entries are single-pass." For Table 7, token counts are reported as "mean generated tokens per problem," but if math token counts are collected under avg@48, it is unclear whether the denominator is a problem, a single rollout, or an averaged rollout. Token counts also depend on decoding parameters, which the paper states are "provider-recommended" and therefore differ across models. For a paper whose central claim is a 3x token-budget advantage, the token-count definition and decoding configuration must be unambiguous and comparable across models.
  4. [Section 5.1 / Section 5.2] Section 5.1 defines the relevant deployment cost as "cost per correct answer: accuracy weighted by the number of tokens generated," but Section 5.2 and the abstract instead emphasize raw token budgets. The paper never reports cost-per-correct-answer values. Because Domyn-Small trails Qwen3.5-9B substantially on AIME 2025 (35.7 vs 90.0) and LiveCodeBench (55.0 vs 86.2), raw token savings alone do not establish the claimed "accuracy-efficiency balance." Please report the cost-per-correct-answer metric explicitly, or soften the claim to a token-budget comparison with accuracy reported separately.
minor comments (5)
  1. [Figure 3] The OLS regression line is fit to only five model points and is labeled a "frontier." This is at best suggestive; consider removing the line or adding a clear caveat that it is not a fitted Pareto frontier.
  2. [Table 7] The table header says "Mean generated tokens per problem," while Figure 3's axis says "Mean avg output tokens per sample." Use consistent terminology and define whether averages are computed over all rollouts or over pass@k samples.
  3. [Section 6.8] The claim of AI Act Article 53 compliance mentions a companion training-data summary artifact, but no URL or appendix pointer is given. Add a reference or state where this artifact is published.
  4. [Section 4.2] The reward function is described with an unusual range notation ("{1,0,−0.8}") and a sign convention for verifier errors. Clarifying the reward computation with a formula would improve reproducibility.
  5. [Abstract] The abstract's "roughly one-third as many tokens as Qwen3.5-9B" is accurate only for the reasoning grand mean in Table 7, not for code or general knowledge. Qualify the claim as "on the reasoning benchmarks evaluated here."

Circularity Check

0 steps flagged

No significant circularity: token-efficiency and accuracy results are measured outputs, not fitted inputs; self-citations are provenance-only.

full rationale

The paper's derivation chain is an empirical pipeline (CPT, SFT, GRPO, DPO, multi-environment GRPO, Sections 3-4) followed by external benchmark measurement (Section 5). The headline token-efficiency numbers (e.g., reasoning grand mean 2,690 vs. Qwen3.5-9B's 8,440) are computed from Table 7 as weighted means of measured tokens per problem; no parameter is fitted to these benchmarks and then reported as a prediction. The only self-citations are provenance statements, e.g., 'The CPT data mix is shared with the pre-training corpus of Domyn Large (Bertolotti et al., 2026)' and the Italia/Colosseum lineage, and they do not carry the central efficiency claim. Two verification gaps are noted but are not circular reductions: Section 5.1 states 'The Domyn-Small results presented here correspond to the latest checkpoint of the multi-environment GRPO stage (Section 4.4); values may be refreshed upon final model release,' while Section 4.4 says 'the released v1.0 weights are the best-validation checkpoint from this run,' so the evaluated checkpoint may differ from the released artifact; and no contamination analysis is provided for the 40+ public instruction datasets. These affect whether the reported numbers describe the downloadable model, not whether the numbers were derived from their own inputs. No uniqueness theorem, fitted-input-renamed-as-prediction, or ansatz-via-citation chain appears.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The central claim depends on a set of hand-chosen training hyperparameters and unstated data assumptions (primarily no contamination check) rather than on a derivation from first principles.

free parameters (5)
  • RoPE base after CPT = 500,000
    Hand-selected to support 32K native context; affects positional encoding and long-context behavior (Section 3.2).
  • GRPO length penalty = 0 to -0.25 at 8192 tokens
    Hand-designed reward component penalizing long generations; directly shapes token efficiency (Section 4.2).
  • Multi-env GRPO KL penalty β = 0.01
    Empirically determined to prevent policy oscillation across task domains (Section 4.4).
  • Multi-env GRPO learning rate = 1.5e-6
    Empirically determined stable configuration after warmup (Section 4.4).
  • DPO β = 0.1
    Standard DPO temperature chosen for the preference stage (Section 4.3).
axioms (4)
  • domain assumption Evaluation benchmarks are not present in the training data.
    The SFT mixture includes 40+ public instruction datasets and CPT includes web-scale data; no contamination analysis is provided (Sections 3.2, 4.1, 5).
  • domain assumption Delta Learning Hypothesis (Geng et al., 2025) holds: preference tuning on weak data yields strong gains.
    The DPO stage is built on this hypothesis and the paper defers to the original paper rather than independently verifying it (Section 4.3).
  • domain assumption Italia 10B is a competent base model as claimed.
    The paper relies on its own unpublished Italia 10B foundation model as the starting point; no formal citation or detailed eval of the base model is provided (Section 3.1).
  • domain assumption Rule-based verifiers used in GRPO correctly score mathematical answers and code execution.
    The reward signal depends on symbolic verifiers and unit tests; any verifier errors propagate into training (Sections 4.2, 4.4).

pith-pipeline@v1.3.0-alltime-deepseek · 28967 in / 14528 out tokens · 122583 ms · 2026-08-02T14:03:07.953244+00:00 · methodology

0 comments
read the original abstract

We introduce Domyn-Small, a 10-billion-parameter open-weight reasoning language model released under the MIT license. Domyn-Small is the product of an initial pre-training phase on 9 trillion tokens multilingual data, followed by a post-training pipeline for reasoning, instruction following, and context extension. For the latter, we performed a Continued Pre-Training (CPT) phase that doubles the native context window to 32K tokens, followed by SFT with a math-focused annealing run. Finally, the RL phase includes GRPO with verifiable rewards, DPO, and a multi-environment GRPO stage spanning five task domains: mathematics, code, multiple-choice QA, instruction-following, and tool calling. The 32K-token native context extends to 128K at inference via YaRN, and a chat-template toggle enables dual-mode reasoning. Against peer models in the 7--10B class (Qwen3.5-9B, OLMo-3-7B-Think, Nemotron-Nano-8B, Ministral-3-8B), Domyn-Small achieves a strong accuracy-efficiency balance: it produces roughly one-third as many tokens as Qwen3.5-9B and approximately 35% of OLMo-3-7B-Think's token budget on core reasoning benchmarks, while delivering strong instruction-following (IFEval 79.9) and competitive science reasoning (GPQA-Diamond 50.0). We release the weights and the post-training recipe alongside Domyn Swarm (Apache~2.0), an open-source framework for scalable LLM inference on HPC clusters developed during this program and used throughout this work.

Figures

Figures reproduced from arXiv: 2607.20448 by Alberto Veneri, Alessandro Rognoni, Andrea Valenti, Dario Salvati, Federico D'Ambrosio, Francesco Bertolotti, Martin Cimmino, Michele Resta, Nicol\`o Ruggeri, Simone Angarano.

Figure 1
Figure 1. Figure 1: High-level training pipeline of Domyn-Small. Starting from the Italia 10 B checkpoint, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Data composition across the 503B-token CPT corpus. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Accuracy/token-budget efficiency frontier across the ten benchmarks listed in Table [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

78 extracted references · 9 linked inside Pith

  1. [1]

    DeepScaleR: Surpassing O1-Preview with a 1.5B Model by Scaling RL , author=

  2. [2]

    GQA : Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints

    Ainslie, Joshua and Lee-Thorp, James and de Jong, Michiel and Zemlyanskiy, Yury and Lebron, Federico and Sanghai, Sumit. GQA : Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023. doi:10.18653/v1/2023.emnlp-main.298

    doi:10.18653/v1/2023.emnlp-main.298 2023

  3. [3]

    2016 , eprint=

    Layer Normalization , author=. 2016 , eprint=

    2016

  4. [4]

    Bowen Peng and Jeffrey Quesnelle and Honglu Fan and Enrico Shippole , booktitle=. Ya. 2024 , url=

    2024

  5. [5]

    So and Wojciech Manke and Hanxiao Liu and Zihang Dai and Noam Shazeer and Quoc V

    David R. So and Wojciech Manke and Hanxiao Liu and Zihang Dai and Noam Shazeer and Quoc V. Le , title =. CoRR , volume =. 2021 , url =. 2109.08668 , timestamp =

    Pith/arXiv arXiv 2021

  6. [6]

    RoFormer: Enhanced transformer with Rotary Position Embedding , journal =

    Jianlin Su and Murtadha Ahmed and Yu Lu and Shengfeng Pan and Wen Bo and Yunfeng Liu , keywords =. RoFormer: Enhanced transformer with Rotary Position Embedding , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.neucom.2023.127063 , url =

    arXiv 2024

  7. [7]

    Attention is All you Need , url =

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N and Kaiser, ukasz and Polosukhin, Illia , booktitle =. Attention is All you Need , url =

  8. [8]

    Kudo, Taku and Richardson, John , booktitle =

  9. [9]

    and Sifre, Laurent , title =

    Hoffmann, Jordan and Borgeaud, Sebastian and Mensch, Arthur and Buchatskaya, Elena and Cai, Trevor and Rutherford, Eliza and de Las Casas, Diego and Hendricks, Lisa Anne and Welbl, Johannes and Clark, Aidan and Hennigan, Tom and Noland, Eric and Millican, Katie and van den Driessche, George and Damoc, Bogdan and Guy, Aurelia and Osindero, Simon and Simony...

    2022

  10. [10]

    Deduplicating Training Data Makes Language Models Better

    Lee, Katherine and Ippolito, Daphne and Nystrom, Andrew and Zhang, Chiyuan and Eck, Douglas and Callison-Burch, Chris and Carlini, Nicholas. Deduplicating Training Data Makes Language Models Better. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022. doi:10.18653/v1/2022.acl-long.577

    doi:10.18653/v1/2022.acl-long.577 2022

  11. [11]

    2022 , eprint=

    Scaling Language Models: Methods, Analysis & Insights from Training Gopher , author=. 2022 , eprint=

    2022

  12. [12]

    Liu , title =

    Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu , title =. Journal of Machine Learning Research , year =

  13. [13]

    CCN et: Extracting High Quality Monolingual Datasets from Web Crawl Data

    Wenzek, Guillaume and Lachaux, Marie-Anne and Conneau, Alexis and Chaudhary, Vishrav and Guzm \'a n, Francisco and Joulin, Armand and Grave, Edouard. CCN et: Extracting High Quality Monolingual Datasets from Web Crawl Data. Proceedings of the Twelfth Language Resources and Evaluation Conference. 2020

    2020

  14. [14]

    Jeffrey Li and Alex Fang and Georgios Smyrnis and Maor Ivgi and Matt Jordan and Samir Yitzhak Gadre and Hritik Bansal and Etash Kumar Guha and Sedrick Keh and Kushal Arora and Saurabh Garg and Rui Xin and Niklas Muennighoff and Reinhard Heckel and Jean Mercat and Mayee F Chen and Suchin Gururangan and Mitchell Wortsman and Alon Albalak and Yonatan Bitton ...

    2024

  15. [15]

    Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research

    Soldaini, Luca and Kinney, Rodney and Bhagia, Akshita and Schwenk, Dustin and Atkinson, David and Authur, Russell and Bogin, Ben and Chandu, Khyathi and Dumas, Jennifer and Elazar, Yanai and Hofmann, Valentin and Jha, Ananya and Kumar, Sachin and Lucy, Li and Lyu, Xinxi and Lambert, Nathan and Magnusson, Ian and Morrison, Jacob and Muennighoff, Niklas and...

    2024

  16. [16]

    2024 , eprint=

    StarCoder 2 and The Stack v2: The Next Generation , author=. 2024 , eprint=

    2024

  17. [17]

    Luca Soldaini and Kyle Lo , year = 2023, title =

    2023

  18. [18]

    The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=

    The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale , author=. The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=

  19. [19]

    Nemotron- CC : Transforming C ommon C rawl into a Refined Long-Horizon Pretraining Dataset

    Su, Dan and Kong, Kezhi and Lin, Ying and Jennings, Joseph and Norick, Brandon and Kliegl, Markus and Patwary, Mostofa and Shoeybi, Mohammad and Catanzaro, Bryan. Nemotron- CC : Transforming C ommon C rawl into a Refined Long-Horizon Pretraining Dataset. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lon...

    doi:10.18653/v1/2025.acl-long.123 2025

  20. [20]

    Effective

    Xiong, Wenhan and Liu, Jingyu and Molybog, Igor and Zhang, Hejia and Bhargava, Prajjwal and Hou, Rui and Martin, Louis and Rungta, Rashi and Sankararaman, Karthik Abinav and Oguz, Barlas and Khabsa, Madian and Fang, Han and Mehdad, Yashar and Narang, Sharan and Malik, Kshitiz and Fan, Angela and Bhosale, Shruti and Edunov, Sergey and Lewis, Mike and Wang,...

    Pith/arXiv arXiv doi:10.48550/arxiv.2309.16039

  21. [21]

    Aakanksha Chowdhery and Sharan Narang and Jacob Devlin and Maarten Bosma and Gaurav Mishra and Adam Roberts and Paul Barham and Hyung Won Chung and Charles Sutton and Sebastian Gehrmann and Parker Schuh and Kensen Shi and Sasha Tsvyashchenko and Joshua Maynez and Abhishek Rao and Parker Barnes and Yi Tay and Noam Shazeer and Vinodkumar Prabhakaran and Emi...

  22. [22]

    Reducing Activation Recomputation in Large Transformer Models , url =

    Korthikanti, Vijay Anand and Casper, Jared and Lym, Sangkug and McAfee, Lawrence and Andersch, Michael and Shoeybi, Mohammad and Catanzaro, Bryan , booktitle =. Reducing Activation Recomputation in Large Transformer Models , url =

  23. [23]

    2020 , eprint=

    Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism , author=. 2020 , eprint=

    2020

  24. [24]

    Back to Basics: Revisiting REINFORCE -Style Optimization for Learning from Human Feedback in LLM s

    Ahmadian, Arash and Cremer, Chris and Gall. Back to Basics: Revisiting REINFORCE -Style Optimization for Learning from Human Feedback in LLM s. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024. doi:10.18653/v1/2024.acl-long.662

    doi:10.18653/v1/2024.acl-long.662 2024

  25. [25]

    2025 , howpublished =

    OLMo-3.1-32B-Instruct , author =. 2025 , howpublished =

    2025

  26. [26]

    2026 , eprint=

    Olmo 3 , author=. 2026 , eprint=

    2026

  27. [27]

    Second Conference on Language Modeling , year=

    The Delta Learning Hypothesis: Preference Tuning on Weak Data can Yield Strong Gains , author=. Second Conference on Language Modeling , year=

  28. [28]

    Thirty-seventh Conference on Neural Information Processing Systems , year=

    Direct Preference Optimization: Your Language Model is Secretly a Reward Model , author=. Thirty-seventh Conference on Neural Information Processing Systems , year=

  29. [29]

    2024 , eprint=

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models , author=. 2024 , eprint=

    2024

  30. [30]

    Guo, Daya and Yang, Dejian and Zhang, Haowei and Song, Junxiao and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Zhang, Ruoyu and Ma, Shirong and Bi, Xiao and Zhang, Xiaokang and Yu, Xingkai and Wu, Yu and Wu, Z. F. and Gou, Zhibin and Shao, Zhihong and Li, Zhuoshu and Gao, Ziyi and Liu, Aixin and Xue, Bing and Wang, Bingxuan and Wu, Bochao and Feng, Bei ...

  31. [31]

    2025 , eprint=

    Qwen3 Technical Report , author=. 2025 , eprint=

    2025

  32. [32]

    Second Conference on Language Modeling , year=

    Tulu 3: Pushing Frontiers in Open Language Model Post-Training , author=. Second Conference on Language Modeling , year=

  33. [33]

    Proceedings of The 27th International Conference on Artificial Intelligence and Statistics , pages =

    A General Theoretical Paradigm to Understand Learning from Human Preferences , author =. Proceedings of The 27th International Conference on Artificial Intelligence and Statistics , pages =. 2024 , editor =

    2024

  34. [34]

    Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles , year=

    Efficient Memory Management for Large Language Model Serving with PagedAttention , author=. Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles , year=

  35. [35]

    2020 , eprint=

    Mastering Complex Control in MOBA Games with Deep Reinforcement Learning , author=. 2020 , eprint=

    2020

  36. [36]

    2023 , eprint=

    Qwen Technical Report , author=. 2023 , eprint=

    2023

  37. [37]

    2020 , eprint=

    Language Models are Few-Shot Learners , author=. 2020 , eprint=

    2020

  38. [38]

    2024 , eprint=

    Gemma: Open Models Based on Gemini Research and Technology , author=. 2024 , eprint=

    2024

  39. [39]

    2023 , eprint=

    Mistral 7B , author=. 2023 , eprint=

    2023

  40. [40]

    2023 , eprint=

    BLOOM: A 176B-Parameter Open-Access Multilingual Language Model , author=. 2023 , eprint=

    2023

  41. [41]

    2023 , eprint=

    LLaMA: Open and Efficient Foundation Language Models , author=. 2023 , eprint=

    2023

  42. [42]

    2023 , eprint=

    Llama 2: Open Foundation and Fine-Tuned Chat Models , author=. 2023 , eprint=

    2023

  43. [43]

    2021 , eprint=

    Program Synthesis with Large Language Models , author=. 2021 , eprint=

    2021

  44. [44]

    2021 , eprint=

    Evaluating Large Language Models Trained on Code , author=. 2021 , eprint=

    2021

  45. [45]

    2018 , eprint=

    Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge , author=. 2018 , eprint=

    2018

  46. [46]

    2021 , eprint=

    Training Verifiers to Solve Math Word Problems , author=. 2021 , eprint=

    2021

  47. [47]

    doi:10.5281/zenodo.12608602 , url =

    Gao, Leo and Tow, Jonathan and Abbasi, Baber and Biderman, Stella and Black, Sid and DiPofi, Anthony and Foster, Charles and Golding, Laurence and Hsu, Jeffrey and Le Noac'h, Alain and Li, Haonan and McDonell, Kyle and Muennighoff, Niklas and Ociepa, Chris and Phang, Jason and Reynolds, Laria and Schoelkopf, Hailey and Skowron, Aviya and Sutawika, Lintang...

    doi:10.5281/zenodo.12608602

  48. [48]

    International Conference on Learning Representations , year=

    Measuring Massive Multitask Language Understanding , author=. International Conference on Learning Representations , year=

  49. [49]

    The Thirteenth International Conference on Learning Representations , year=

    LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code , author=. The Thirteenth International Conference on Learning Representations , year=

  50. [50]

    Challenging BIG -Bench Tasks and Whether Chain-of-Thought Can Solve Them

    Suzgun, Mirac and Scales, Nathan and Sch. Challenging BIG -Bench Tasks and Whether Chain-of-Thought Can Solve Them. Findings of the Association for Computational Linguistics: ACL 2023. 2023. doi:10.18653/v1/2023.findings-acl.824

    doi:10.18653/v1/2023.findings-acl.824 2023

  51. [51]

    H ella S wag: Can a Machine Really Finish Your Sentence?

    Zellers, Rowan and Holtzman, Ari and Bisk, Yonatan and Farhadi, Ali and Choi, Yejin. H ella S wag: Can a Machine Really Finish Your Sentence?. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019. doi:10.18653/v1/P19-1472

    doi:10.18653/v1/p19-1472 2019

  52. [52]

    Measuring Mathematical Problem Solving With the

    Dan Hendrycks and Collin Burns and Saurav Kadavath and Akul Arora and Steven Basart and Eric Tang and Dawn Song and Jacob Steinhardt , booktitle=. Measuring Mathematical Problem Solving With the. 2021 , url=

    2021

  53. [53]

    Approximating

    Schulman, John , year =. Approximating

  54. [54]

    Bowman , booktitle=

    David Rein and Betty Li Hou and Asa Cooper Stickland and Jackson Petty and Richard Yuanzhe Pang and Julien Dirani and Julian Michael and Samuel R. Bowman , booktitle=. 2024 , url=

    2024

  55. [55]

    MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark , url =

    Wang, Yubo and Ma, Xueguang and Zhang, Ge and Ni, Yuansheng and Chandra, Abhranil and Guo, Shiguang and Ren, Weiming and Arulraj, Aaran and He, Xuan and Jiang, Ziyan and Li, Tianle and Ku, Max and Wang, Kai and Zhuang, Alex and Fan, Rongqi and Yue, Xiang and Chen, Wenhu , booktitle =. MMLU-Pro: A More Robust and Challenging Multi-Task Language Understandi...

    doi:10.52202/079017-3018

  56. [56]

    2023 , eprint=

    Instruction-Following Evaluation for Large Language Models , author=. 2023 , eprint=

    2023

  57. [57]

    2024 , url=

    Cheng-Ping Hsieh and Simeng Sun and Samuel Kriman and Shantanu Acharya and Dima Rekesh and Fei Jia and Boris Ginsburg , booktitle=. 2024 , url=

    2024

  58. [58]

    Gonzalez , booktitle=

    Shishir G Patil and Huanzhi Mao and Fanjia Yan and Charlie Cheng-Jie Ji and Vishnu Suresh and Ion Stoica and Joseph E. Gonzalez , booktitle=. The Berkeley Function Calling Leaderboard (. 2025 , url=

    2025

  59. [59]

    2026 , month = feb, howpublished =

    2026

  60. [60]

    2025 , howpublished =

    2025

  61. [61]

    2025 , howpublished =

    Ministral-3-8B-Reasoning-2512 , author =. 2025 , howpublished =

    2025

  62. [62]

    2022 , eprint =

    Language Models are Multilingual Chain-of-Thought Reasoners , author =. 2022 , eprint =

    2022

  63. [63]

    2025 , eprint=

    Magistral , author=. 2025 , eprint=

    2025

  64. [64]

    2025 , eprint=

    Llama-Nemotron: Efficient Reasoning Models , author=. 2025 , eprint=

    2025

  65. [65]

    2026 , url=

    Qiying Yu and Zheng Zhang and Ruofei Zhu and Yufeng Yuan and Xiaochen Zuo and YuYue and Weinan Dai and Tiantian Fan and Gaohong Liu and Juncai Liu and LingJun Liu and Xin Liu and Haibin Lin and Zhiqi Lin and Bole Ma and Guangming Sheng and Yuxuan Tong and Chi Zhang and Mofan Zhang and Ru Zhang and Wang Zhang and Hang Zhu and Jinhua Zhu and Jiaze Chen and ...

    2026

  66. [66]

    2025 , note =

    NeMo RL: A Scalable and Efficient Post-Training Library , author =. 2025 , note =

    2025

  67. [67]

    Domyn-Large: The Journey of a European Sovereign

    Bertolotti, Francesco and Veneri, Alberto and Rognoni, Alessandro and D'Ambrosio, Federico and Morais, Carlos and Perez, Sergio and Pilzer, Andrea and Cimmino, Martin , year =. Domyn-Large: The Journey of a European Sovereign

  68. [68]

    2024 , howpublished =

    Continued Pretraining of State-of-the-Art. 2024 , howpublished =

    2024

  69. [69]

    , booktitle =

    Parrish, Alicia and Chen, Angelica and Nangia, Nikita and Padmakumar, Vishakh and Phang, Jason and Thompson, Jana and Htut, Phu Mon and Bowman, Samuel R. , booktitle =. 2022 , url =. 2110.08193 , archivePrefix =

    Pith/arXiv arXiv 2022

  70. [70]

    and Dombrowski, Ann-Kathrin and Goel, Shashwat and Phan, Long and others , booktitle =

    Li, Nathaniel and Pan, Alexander and Gopal, Anjali and Yue, Summer and Berrios, Daniel and Gatti, Alice and Li, Justin D. and Dombrowski, Ann-Kathrin and Goel, Shashwat and Phan, Long and others , booktitle =. The. 2024 , url =. 2403.03218 , archivePrefix =

    Pith/arXiv arXiv 2024

  71. [71]

    2021 , url =

    Nadeem, Moin and Bethke, Anna and Reddy, Siva , booktitle =. 2021 , url =. 2004.09456 , archivePrefix =

    Pith/arXiv arXiv 2021

  72. [72]

    Souly, Alexandra and Lu, Qingyuan and Bowen, Dillon and Trinh, Tu and Hsieh, Elvis and Pandey, Sana and Abbeel, Pieter and Svegliato, Justin and Emmons, Scott and Watkins, Olivia and Toyer, Sam , year =. A. 2402.10260 , archivePrefix =

    Pith/arXiv arXiv

  73. [73]

    Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics , year =

    R. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics , year =. 2308.01263 , archivePrefix =

    Pith/arXiv arXiv 2024

  74. [74]

    2402.07688 , archivePrefix =

    Tihanyi, Norbert and Ferrag, Mohamed Amine and Jain, Ridhi and Bisztray, Tamas and Debbah, Merouane , year =. 2402.07688 , archivePrefix =

    Pith/arXiv arXiv

  75. [75]

    2312.15838 , archivePrefix =

    Liu, Zefang , year =. 2312.15838 , archivePrefix =

    Pith/arXiv arXiv

  76. [76]

    2024 , howpublished =

    Inspect. 2024 , howpublished =

    2024

  77. [77]

    Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations , year =

    2024

  78. [78]

    doi:10.17815/jlsrf-8-186 , abstract =

    Journal of large-scale research facilities JLSRF , author =. doi:10.17815/jlsrf-8-186 , abstract =

    doi:10.17815/jlsrf-8-186