Pith. sign in

REVIEW 3 major objections 4 minor 66 references

NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that freezing language-sensitive neurons while updating vision-adaptive ones during multimodal instruction tuning recovers 94.5% of lost language ability without sacrificing multimodal performance.

desk verdict A solid, broad, and useful method for language-preserving multimodal tuning; the causal neuron-role claim outruns the evidence, and the reporting needs tightening. read the letter →

arxiv 2608.08107 v1 pith:DGZI73QF submitted 2026-08-08 cs.CL cs.AI

Jiayue Jin , Jingwei Zhang , Chen Wang , Jing Liu , Longteng Guo This is my paper · ORCID
classification cs.CLcs.AI
keywords multimodallargelanguagemodelsforgettingneuronplasticityimportancemodalitypreferenceinstructiontuningupdateallocationcatastrophic
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Turning a pretrained language model into a multimodal model by visual instruction tuning usually costs pure-text reasoning ability: the paper measures a roughly five-point drop over 11 language benchmarks on its default backbone. NeuPAT is a training-time method that treats this degradation as a neuron-allocation problem rather than a data or architecture problem. Before tuning, it probes the model with text-only and visual samples and scores every neuron by how strongly it responds to each modality, then freezes language-sensitive neurons, fully updates vision-adaptive and low-importance neurons, and only regularizes neurons important to both. On the 11 language benchmarks the paper reports recovering the large majority of the lost capability, with the average climbing from 43.92 to 49.06 and slightly surpassing the original LLM's 48.99, while the multimodal average stays at 61.05 versus 60.98. The claim is that heterogeneous neuron plasticity is a general, architecture-agnostic resource that can be allocated explicitly.

What carries the argument

The load-bearing object is the modality-associated importance score $s_{l,u}(\mathcal{D}) = \|W_{l,u}^{\mathrm{out}}\|_F \cdot \mathrm{RMS}_{x\in\mathcal{D}, t\in V(x)} \|h_{l,u}(x_t)\|_2$, an activation-aware estimate of how much a neuron writes to the residual stream under a given input modality. From its text and vision variants the paper forms a modality preference $P_{l,u}=v_{l,u}-t_{l,u}$ and an overall importance $I_{l,u}=(v_{l,u}+t_{l,u})/2$, then for each layer selects the smallest neuron sets whose cumulative importance reaches a coverage threshold of 0.8. The four role sets (language, multimodal, shared, reserve) determine the update rules, with shared neurons regularized by an input-side L2 penalty plus an output-side cosine penalty; this mapping from measured response to role assignment is what converts an observation about plasticity into a training algorithm.

What would settle it

Take a tuned NeuPAT model and ablate the top-ranked language-associated neurons on reasoning benchmarks such as BBH and GSM8K; if those benchmarks stay roughly unchanged, the identified neurons are not causal for preservation. Alternatively, invert the importance ranking during training, freezing what the probe calls vision neurons and updating what it calls language neurons, and check whether the language average falls back toward the vanilla-tuned level; the ranking, not the overall freeze-and-update regimen, is what the preservation claim predicts.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a pretrained LLM backbone does not adapt uniformly when visual instruction tuning is applied; neurons divide into adaptively distinct populations. A neuron's modality preference is estimated by an importance score that combines its output-side weight norm with its activation magnitude under probing inputs: neurons whose importance is dominated by text are language-associated, those dominated by vision are multimodal-adaptive, those important to both are shared, and those important to neither are reserve. NeuPAT freezes the language-associated neurons, fully trains the multimodal-adaptive and reserve neurons, and regularizes shared neurons with an input-side L2 penalty plus an output-side cosine constraint, so the residual-stream direction of shared neurons cannot drift far from pretrained behavior. The paper argues this role-aware allocation is why, across six LLMs of different families and sizes, the 11-benchmark language average is preserved (49.06 versus the original LLM's 48.99) while the 5-benchmark multimodal average holds at 61.05 versus vanilla tuning's 60.98.

Load-bearing premise

The load-bearing premise is that the importance score from Eq. (1) identifies neurons whose preservation actually protects language ability; the paper's own Section 8 calls the allocation a response-based functional approximation rather than a causal explanation, so if that proxy identifies only which inputs a neuron tends to see, freezing the wrong neurons could hurt both text and vision performance.

Editorial extensions

If this is right

  • Multimodal instruction tuning can be made language-preserving without text replay, extra data, or architectural changes: the same recipe, probe, assign roles, and constrain updates, is the intervention.
  • The method is architecture-agnostic in the sense that the probing and allocation steps read only forward activations and weight slices, so they should transfer to new LLM families as long as the importance estimate is recomputed for each backbone.
  • Because the probing set is small and used only for forward passes, the overhead before training is a few hundred forward passes rather than a second optimization loop.
  • The ablations imply that the balance is load-bearing: freezing vision neurons or updating language neurons each erodes one side of the text-multimodal trade-off, so the four-role split, not just any freeze, is what preserves language.
  • Recovering the original LLM's language average while holding multimodal performance means practitioners could expand an LLM into an MLLM without budgeting for a separate language-preservation stage.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the same probing-plus-allocation recipe could be re-targeted at other capability axes, such as coding, instruction following, or safety, by choosing probing sets that isolate those skills; the paper does not test this.
  • Beyond the paper, the main comparison set is dominated by methods that intervene after training or with extra supervision, so a head-to-head against text-replay at matched compute would show whether neuron allocation and data replay are substitutes, complements, or partly redundant.
  • Beyond the paper, a causal test of the mechanism would be to measure language benchmarks after ablating the top language-associated neurons in the tuned model; a sharp drop would show the identified neurons are load-bearing, while no drop would suggest the preservation comes from elsewhere in the procedure.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes NeuPAT, a neuron-level plasticity allocation method for multimodal instruction tuning that aims to preserve the language capabilities of a pretrained LLM while still adapting it to visual instruction data. NeuPAT first computes modality-associated importance scores for each neuron as the product of its output weight norm and its activation RMS on small text-only and vision probing sets, then partitions neurons into language-associated, multimodal-adaptive, shared, and reserve roles via cumulative importance thresholds. During tuning, language-associated neurons are frozen, multimodal-adaptive and reserve neurons receive full updates, and shared neurons receive input-side L2 and output-side cosine regularization. Experiments compare NeuPAT with vanilla tuning, LoRA, EWC, WINGS, TIES, Locate-then-Merge, and PlaM across six LLM backbones, eleven text benchmarks, and five multimodal benchmarks, reporting that NeuPAT recovers most of the language degradation caused by vanilla tuning while maintaining comparable multimodal performance. Appendices provide sensitivity analyses over the importance threshold, probing set size, and regularization coefficients, plus complete cross-backbone and ablation tables.

Significance. If the proposed mechanism is validated, NeuPAT would be a practical, architecture-agnostic alternative to text replay, architectural modification, and post-hoc merging for language-preserving multimodal expansion. The paper has notable strengths: broad empirical coverage across six model families and scales, role-wise ablation studies, sensitivity analyses over all free hyperparameters, and complete results in the appendix. The core limitation is that the central text-versus-vision role distinction rests on an importance proxy whose causal validity is not tested, and the headline recovery numbers are internally inconsistent. With additional ablations that vary the score formula itself and more careful reporting of run-to-run variability, the contribution would be substantially stronger.

major comments (3)
  1. [Abstract; §1; §5.2.1, Table 1] The headline recovery statistic is internally inconsistent. The abstract states that NeuPAT recovers 94.5% of language capability degradation on 11 language benchmarks, the introduction reports 90.0% on 4 language reasoning benchmarks, and Table 1 yields 5.14/5.07 ≈ 101.4% recovery, with NeuPAT slightly surpassing the original LLM average (49.06 vs. 48.99). These numbers are not equivalent, and no recovery formula is defined. Please define the recovery metric explicitly, apply it consistently to the full benchmark set, and report the resulting value for all claims.
  2. [§4.1–4.3; §5.2.3, Tables 4–6; §8] The central claim that the neurons identified by Eq. (1) play distinct text-preserving versus vision-adaptive roles is not yet supported by the ablations. Tables 4–6 vary the allocation strategy and the update constraint while keeping the importance score formula fixed; they never vary the score formula itself. I request ablations that (i) omit the output-weight-norm term, (ii) omit the activation-RMS term, (iii) swap or shuffle the text and vision scores at matched set sizes, and (iv) freeze the vision-preferred set instead of the text-preferred set. Without such ablations, the results are consistent with the weaker explanation that any well-chosen high-magnitude subset of neurons can be frozen to preserve language, rather than with the claimed modality-specific mechanism. The concession in Section 8 that the allocation is a response-based functional approximation should be reflected in the abstract and conclusions unless these ablations are added.
  3. [§5.2; Tables 1–3, 8–10] All experimental results are reported as single runs without error bars or significance tests, and the default hyperparameters τ_a = 0.8 and λ_in = λ_out = 0.1 are selected from sensitivity sweeps on the same evaluation benchmarks used for the headline comparisons. Because one headline claim is a 0.07-point average 'surpass' of the original LLM, run-to-run noise could change the conclusion. Please report multiple seeds or at least per-benchmark variance for the main tables, and either select hyperparameters on a validation split or explicitly state that the sensitivity analysis doubles as test-set model selection.
minor comments (4)
  1. [Algorithm 1; §4.2] Notation is inconsistent: Algorithm 1 uses C^{text}, C^{vision}, C^{high}, and C^{low}, while Section 4.2 and Table 4 use C^{lang}, C^{multi}, C^{shared}, and C^{reserve}. Please unify the notation throughout.
  2. [Eq. (6)] The definition of C^{reserve} is ambiguous in the inline typesetting: it should be the complement of T_l ∪ V_l with respect to all neurons in the layer, but the current rendering can be read as the union itself. Please rewrite the equation with an explicit complement operator.
  3. [§4.1] The sentence 'the importance scores are normalized using layer-wise normalization using Eq. 2' should reference Eq. (3), because Eq. (2) defines only the raw scores s^V and s^T, not the normalized scores v and t.
  4. [Tables 2 and 3] The multimodal benchmark is abbreviated as 'MMS' in Table 2 but as 'MMStar' in Table 3 and Section 5.1.3; please use a single abbreviation.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity; NeuPAT's language-preservation claim is empirically validated against external benchmarks and disjoint probes.

full rationale

The central claim is empirical rather than derivational. Equation (1) is an importance proxy taken from external work (Sun et al., 2024); the text and vision probing sets are disjoint from all evaluation benchmarks (Appendix C.1); and the role allocation (Eqs. 5-6) is tested against Random and Fixed-ratio partitions (Table 6) and per-role ablations (Table 4). Freezing language-associated neurons could plausibly have failed to recover language scores, and the fact that it did not is evidence rather than identity. The paper's Section 8 disclaimer that the allocation is 'response-based functional approximations rather than causal explanations of model behavior' is a limitation, not a circularity. The only mild concern is hyperparameter justification via sensitivity tables on the same benchmarks, but the paper states defaults were fixed before sensitivity analyses ('All sensitivity analyses are conducted after fixing the default configuration used in the main experiments'), so this is at most a reporting nuance, not a construction-level reduction. No self-citation chain or imported uniqueness theorem is load-bearing.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The method's free parameters (tau, lambdas, probe size) are tuned on the evaluation benchmarks, which inflates the apparent gains. The axioms are the unverified proxy assumption about importance scores and the representativeness of the probing sets.

free parameters (3)
  • target importance mass tau = 0.8
    Selected via sensitivity analysis on the evaluation benchmarks (Table 8); controls the size of language- and vision-critical neuron sets.
  • shared-neuron regularization coefficients lambda_in, lambda_out = 0.1
    Chosen via sensitivity analysis (Table 10) based on text/multimodal trade-off; higher values reduce multimodal performance.
  • probing set size N_a = 2048
    Selected via sensitivity analysis (Table 9); performance is relatively stable across sizes, with 2048 giving the best averages.
assumptions (3)
  • domain assumption Activation-based importance score (Eq. 1) is a valid proxy for neuron importance and plasticity.
    The method relies on output weight norm times activation magnitude to identify language- and vision-critical neurons; stated as 'inspired by activation-aware importance estimation' (Section 3.1) but not causally validated.
  • domain assumption Probing sets are representative of the modality distributions and disjoint from evaluation benchmarks.
    Text probing uses CodeAlpaca, MetaMathQA, dolly, HaluEval; vision probing uses LLaVA-NeXT-780K training data. Claims disjointness but does not demonstrate it mechanically.
  • domain assumption Freezing language-associated neurons and regularizing shared neurons will not impair multimodal adaptation.
    Central to the method; validated only empirically through ablations, not by theory.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs." pith.science (2026). https://pith.science/paper/DGZI73QF

@misc{pith2026260808107,
  author       = {Pith},
  title        = {Pith review of: NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DGZI73QF}},
  note         = {Machine review of arXiv:2608.08107}
}
read the original abstract

Multimodal expansion of large language models (LLMs) enables new perceptual capabilities but often compromises the language intelligence acquired during pretraining. In this work, we investigate this phenomenon from the perspective of internal adaptation dynamics and discover that neurons in pretrained LLMs exhibit heterogeneous plasticity during multimodal learning: some neurons are critical for preserving language capabilities, while others are more adaptive to multimodal knowledge. Based on this insight, we propose NeuPAT (Neuron-aware Plasticity Allocation Tuning), a lightweight and architecture-agnostic framework that allocates neuron-wise update constraints during multimodal instruction tuning. NeuPAT uses a small-scale probing stage to estimate neuron adaptation patterns and selectively protects language-sensitive neurons while promoting multimodal adaptation through more plastic neurons. Experiments across diverse LLM families demonstrate that NeuPAT recovers 94.5\% of the language capability degradation caused by vanilla tuning on 11 language benchmarks while maintaining comparable multimodal performance, providing an effective approach for capability-preserving multimodal expansion.

Figures

Figures reproduced from arXiv: 2608.08107 by the authors.

Figure 1
Figure 1. Pure-text capability degradation and re￾covery after visual instruction tuning. (a) Expanding an LLM into an MLLM with multimodal instruction data can induce language forgetting. (b) Vanilla tun￾ing degrades performance on representative pure-text benchmarks, whereas NeuPAT recovers most of the lost capability. Scores are normalized by the original LLM performance on each benchmark. it to modality-specific encoders … view at source ↗
Figure 2
Figure 2. Neuron modality preference and importance during multimodal expansion of LLMs. (a) Raw text- and vision-associated importance scores, showing heterogeneous modality associations across neurons. (b) Layer-wise modality preference Pl,u = vl,u − tl,u, with neurons sorted by preference within each layer. (c) Overall importance Il,u = (vl,u + tl,u)/2 under the same ordering, showing diverse importance levels among neuron… view at source ↗
Figure 3
Figure 3. Overview of NeuPAT. NeuPAT identifies neuron adaptation roles through lightweight modality probing and assigns role-specific update constraints during multimodal tuning, preserving language intelligence while acquiring multimodal capabilities. We compute the visual- and language-associated importance scores as s V l,u = sl,u(DV ), sT l,u = sl,u(DT ). (2) For visualization and comparison within each layer, the scores… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (a), the resulting neuron map is closely consistent with the modality-associated response patterns observed in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Layer-wise neuron distributions across LLM backbones. High response neurons form the largest group in most layers, while the others vary across models and depths. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 12 canonical work pages

  1. [1]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  2. [2]

    Advances in neural information processing systems , volume=

    Instructblip: Towards general-purpose vision-language models with instruction tuning , author=. Advances in neural information processing systems , volume=

  3. [3]

    arXiv preprint arXiv:2409.12191 , year=

    Qwen2-vl: Enhancing vision-language model's perception of the world at any resolution , author=. arXiv preprint arXiv:2409.12191 , year=

    arXiv

  4. [4]

    arXiv preprint arXiv:2511.21631 , year=

    Qwen3-vl technical report , author=. arXiv preprint arXiv:2511.21631 , year=

    arXiv

  5. [5]

    arXiv preprint arXiv:2408.03326 , year=

    Llava-onevision: Easy visual task transfer , author=. arXiv preprint arXiv:2408.03326 , year=

    arXiv

  6. [6]

    Science China Information Sciences , volume=

    MNAFT: modality neuron-aware fine-tuning of multimodal large language models for image translation , author=. Science China Information Sciences , volume=. 2026 , publisher=

    work page 2026

  7. [7]

    arXiv preprint arXiv:2504.10479 , year=

    Internvl3: Exploring advanced training and test-time recipes for open-source multimodal models , author=. arXiv preprint arXiv:2504.10479 , year=

    arXiv

  8. [8]

    arXiv preprint arXiv:2504.07491 , year=

    Kimi-vl technical report , author=. arXiv preprint arXiv:2504.07491 , year=

    arXiv

Show all 66 references
  1. [9]

    5: Advancing open-source multimodal models in versatility, reasoning, and efficiency , author=

    Internvl3. 5: Advancing open-source multimodal models in versatility, reasoning, and efficiency , author=. arXiv preprint arXiv:2508.18265 , year=

    arXiv

  2. [10]

    Advances in Neural Information Processing Systems , volume=

    Wings: Learning multimodal llms without text-only forgetting , author=. Advances in Neural Information Processing Systems , volume=

  3. [11]

    arXiv preprint arXiv:2412.03467 , year=

    Training-free mitigation of language reasoning degradation after multimodal instruction tuning , author=. arXiv preprint arXiv:2412.03467 , year=

    arXiv

  4. [12]

    arXiv preprint arXiv:2403.05525 , year=

    Deepseek-vl: towards real-world vision-language understanding , author=. arXiv preprint arXiv:2403.05525 , year=

    arXiv

  5. [13]

    , author=

    Lora: Low-rank adaptation of large language models. , author=. Iclr , volume=

  6. [14]

    Advances in neural information processing systems , volume=

    Language models are few-shot learners , author=. Advances in neural information processing systems , volume=

  7. [15]

    Journal of machine learning research , volume=

    Palm: Scaling language modeling with pathways , author=. Journal of machine learning research , volume=

  8. [16]

    arXiv preprint arXiv:2303.08774 , year=

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    arXiv

  9. [17]

    arXiv preprint arXiv:2302.13971 , year=

    Llama: Open and efficient foundation language models , author=. arXiv preprint arXiv:2302.13971 , year=

    arXiv

  10. [18]

    European conference on computer vision , pages=

    Mmbench: Is your multi-modal model an all-around player? , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  11. [19]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Mmmu: A massive multi-discipline multimodal understanding and reasoning benchmark for expert agi , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  12. [20]

    Advances in Neural Information Processing Systems , volume=

    Are we on the right way for evaluating large vision-language models? , author=. Advances in Neural Information Processing Systems , volume=

  13. [21]

    Proceedings of the national academy of sciences , volume=

    Overcoming catastrophic forgetting in neural networks , author=. Proceedings of the national academy of sciences , volume=. 2017 , publisher=

    2017

  14. [22]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Iaa: Inner-adaptor architecture empowers frozen large language model with multimodal capabilities , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  15. [23]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    GenieBlue: Integrating both Linguistic and Multimodal Capabilities for Large Language Models on Mobile Devices , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  16. [24]

    arXiv preprint arXiv:2505.16703 , year=

    Locate-then-Merge: Neuron-Level Parameter Fusion for Mitigating Catastrophic Forgetting in Multimodal LLMs , author=. arXiv preprint arXiv:2505.16703 , year=

    arXiv

  17. [25]

    arXiv preprint arXiv:2601.07645 , year=

    PlaM: Training-Free Plateau-Guided Model Merging for Better Visual Grounding in MLLMs , author=. arXiv preprint arXiv:2601.07645 , year=

  18. [26]

    CCF International Conference on Natural Language Processing and Chinese Computing , pages=

    DPIMerge: An Efficient Dynamic Parameter Interpolation Framework for Alleviating Pure Text Forgetting in Multimodal Large Models , author=. CCF International Conference on Natural Language Processing and Chinese Computing , pages=. 2025 , organization=

    2025

  19. [27]

    arXiv preprint arXiv:2204.14198 , year=

    Flamingo: a visual language model for few-shot learning , author=. arXiv preprint arXiv:2204.14198 , year=

    arXiv

  20. [28]

    International conference on machine learning , pages=

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models , author=. International conference on machine learning , pages=. 2023 , organization=

    2023

  21. [29]

    International Conference on Learning Representations , volume=

    Minigpt-4: Enhancing vision-language understanding with advanced large language models , author=. International Conference on Learning Representations , volume=

  22. [30]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Improved baselines with visual instruction tuning , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  23. [31]

    arXiv preprint arXiv:2309.16609 , year=

    Qwen technical report , author=. arXiv preprint arXiv:2309.16609 , year=

    arXiv

  24. [32]

    arXiv preprint arXiv:2606.17057 , year=

    Correct When Paired, Wrong When Split: Decoupling and Editing Modality-Specific Neurons in MLLMs , author=. arXiv preprint arXiv:2606.17057 , year=

    arXiv

  25. [33]

    arXiv preprint arXiv:2601.03115 , year=

    Discovering and Causally Validating Emotion-Sensitive Neurons in Large Audio-Language Models , author=. arXiv preprint arXiv:2601.03115 , year=

    arXiv

  26. [34]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Modality-aware neuron pruning for unlearning in multimodal large language models , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  27. [35]

    arXiv preprint arXiv:2510.10238 , year=

    The Achilles' Heel of LLMs: How Altering a Handful of Neurons Can Cripple Language Abilities , author=. arXiv preprint arXiv:2510.10238 , year=

  28. [36]

    International Conference on Learning Representations , volume=

    A simple and effective pruning approach for large language models , author=. International Conference on Learning Representations , volume=

  29. [37]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Region-based cluster discrimination for visual representation learning , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  30. [38]

    arXiv preprint arXiv:2503.01743 , year=

    Phi-4-mini technical report: Compact yet powerful multimodal language models via mixture-of-loras , author=. arXiv preprint arXiv:2503.01743 , year=

    arXiv

  31. [39]

    arXiv preprint arXiv:2505.09388 , year=

    Qwen3 technical report , author=. arXiv preprint arXiv:2505.09388 , year=

    arXiv

  32. [40]

    arXiv preprint arXiv:2412.15115 , year =

    Qwen2.5 Technical Report , author =. arXiv preprint arXiv:2412.15115 , year =

    arXiv

  33. [41]

    arXiv preprint arXiv:2407.21783 , year=

    The llama 3 herd of models , author=. arXiv preprint arXiv:2407.21783 , year=

    arXiv

  34. [42]

    LLaVA-NeXT: Improved reasoning, OCR, and world knowledge , url=

    Liu, Haotian and Li, Chunyuan and Li, Yuheng and Li, Bo and Zhang, Yuanhan and Shen, Sheng and Lee, Yong Jae , month=. LLaVA-NeXT: Improved reasoning, OCR, and world knowledge , url=

  35. [43]

    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...

    doi

  36. [44]

    Findings of the Association for Computational Linguistics: NAACL 2025 , pages=

    Lmms-eval: Reality check on the evaluation of large multimodal models , author=. Findings of the Association for Computational Linguistics: NAACL 2025 , pages=

    2025

  37. [45]

    Social IQa: Commonsense reasoning about social interactions , author=. Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) , pages=

    2019

  38. [46]

    and Ernest Davis and Leora Morgenstern

    Levesque, Hector J. and Ernest Davis and Leora Morgenstern. The winograd schema challenge. 13th International Conference on the Principles of Knowledge Representation and Reasoning, KR 2012. 2012

    2012

  39. [47]

    arXiv preprint arXiv:2110.14168 , year=

    Training verifiers to solve math word problems , author=. arXiv preprint arXiv:2110.14168 , year=

    arXiv

  40. [48]

    arXiv preprint arXiv:2311.12022 , year=

    Gpqa: A graduate-level google-proof q&a benchmark , author=. arXiv preprint arXiv:2311.12022 , year=

    arXiv

  41. [49]

    DROP: A reading comprehension benchmark requiring discrete reasoning over paragraphs , author=. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages=

    2019

  42. [50]

    Transactions of the Association for Computational Linguistics , volume=

    Coqa: A conversational question answering challenge , author=. Transactions of the Association for Computational Linguistics , volume=. 2019 , publisher=

    2019

  43. [51]

    Findings of the Association for Computational Linguistics: ACL 2023 , pages=

    Challenging big-bench tasks and whether chain-of-thought can solve them , author=. Findings of the Association for Computational Linguistics: ACL 2023 , pages=

    2023

  44. [52]

    arXiv preprint arXiv:1502.05698 , year=

    Towards ai-complete question answering: A set of prerequisite toy tasks , author=. arXiv preprint arXiv:1502.05698 , year=

    arXiv

  45. [53]

    arXiv preprint arXiv:1803.05457 , year=

    Think you have solved question answering? try arc, the ai2 reasoning challenge , author=. arXiv preprint arXiv:1803.05457 , year=

    arXiv

  46. [54]

    arXiv preprint arXiv:2410.05669 , year=

    Acpbench: Reasoning about action, change, and planning , author=. arXiv preprint arXiv:2410.05669 , year=

  47. [55]

    LogiQA 2.0—An Improved Dataset for Logical Reasoning in Natural Language Understanding , year=

    Liu, Hanmeng and Liu, Jian and Cui, Leyang and Teng, Zhiyang and Duan, Nan and Zhou, Ming and Zhang, Yue , journal=. LogiQA 2.0—An Improved Dataset for Logical Reasoning in Natural Language Understanding , year=

  48. [56]

    arXiv preprint arXiv:2108.07732 , year=

    Program synthesis with large language models , author=. arXiv preprint arXiv:2108.07732 , year=

    arXiv

  49. [57]

    arXiv preprint arXiv:2103.03874 , year=

    Measuring mathematical problem solving with the math dataset , author=. arXiv preprint arXiv:2103.03874 , year=

    arXiv

  50. [58]

    arXiv preprint arXiv:2009.03300 , year=

    Measuring massive multitask language understanding , author=. arXiv preprint arXiv:2009.03300 , year=

    2009 arXiv

  51. [59]

    Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers) , pages=

    Truthfulqa: Measuring how models mimic human falsehoods , author=. Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers) , pages=

  52. [60]

    Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages=

    A dataset of information-seeking questions and answers anchored in research papers , author=. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages=

    2021

  53. [61]

    Proceedings of the 2023 conference on empirical methods in natural language processing , pages=

    Evaluating object hallucination in large vision-language models , author=. Proceedings of the 2023 conference on empirical methods in natural language processing , pages=

    2023

  54. [62]

    2024 , month = apr, howpublished =

    Grok-1.5 Vision Preview , author =. 2024 , month = apr, howpublished =

    2024

  55. [63]

    GitHub repository , howpublished =

    Sahil Chaudhary , title =. GitHub repository , howpublished =. 2023 , publisher =

    2023

  56. [64]

    arXiv preprint arXiv:2309.12284 , year=

    Metamath: Bootstrap your own mathematical questions for large language models , author=. arXiv preprint arXiv:2309.12284 , year=

    arXiv

  57. [65]

    2023 , url =

    Mike Conover and Matt Hayes and Ankit Mathur and Jianwei Xie and Jun Wan and Sam Shah and Ali Ghodsi and Patrick Wendell and Matei Zaharia and Reynold Xin , title =. 2023 , url =

    2023

  58. [66]

    The 2023 Conference on Empirical Methods in Natural Language Processing , year=

    Halueval: A large-scale hallucination evaluation benchmark for large language models , author=. The 2023 Conference on Empirical Methods in Natural Language Processing , year=

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.