Papers with downstream performance
Taming Extreme Tokens: Covariance-Aware GRPO with Gaussian-Kernel Advantage Reweighting (2026.acl-short)
Copied to clipboard
| Challenge: | Excessive exploitation can cause the model to become overconfident in its suboptimal solutions, thereby limiting its capabilities to explore novel reasoning strategies. |
| Approach: | They propose a method that dynamically down-weights extreme token-level updates via a Gaussian kernel and reduces the instability caused by the trade-off. |
| Outcome: | The proposed method improves downstream performance across reasoning benchmarks and stabilizes entropy as training progresses. |
Best-First Beam Search (2020.tacl-1)
Copied to clipboard
| Challenge: | Currently, beam search is the default for decoding structured predictors . however, little work has been done to speed up beam search itself . |
| Approach: | They propose a beam search algorithm that prunes the scoring function to a monotonic sequence length, which allows for safe pruning of hypotheses that cannot be in the final set of hypothecies. |
| Outcome: | The proposed method can be implemented up to 10x faster in practice. |
ScalingFilter: Assessing Data Quality through Inverse Utilization of Scaling Laws (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing quality filtering methods rely on a high-quality dataset as reference . Existing methods introduce potential biases and compromise diversity . |
| Approach: | They propose a method that evaluates text quality based on the perplexity difference between two language models trained on the same data. |
| Outcome: | The proposed approach improves performance of pre-trained models without increasing training costs. |
MANTa: Efficient Gradient-Based Tokenization for End-to-End Robust Language Modeling (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Subword tokenization algorithms have been an essential component of language modeling but their static nature results in important flaws that degrade the models’ downstream performance and robustness. |
| Approach: | They propose a module for Adaptive Neural TokenizAtion that is differentiable and trained end-to-end with the language model. |
| Outcome: | The proposed tokenizer improves robustness to character perturbations and out-of-domain data. |
Tokenizer Choice For LLM Training: Negligible or Crucial? (2024.findings-naacl)
Copied to clipboard
Mehdi Ali, Michael Fromm, Klaudia Thellmann, Richard Rutmann, Max Lübbering, Johannes Leveling, Katrin Klug, Jan Ebert, Niclas Doll, Jasper Buschhoff, Charvi Jain, Alexander Weber, Lena Jurkschat, Hammam Abdelwahab, Chelsea John, Pedro Ortiz Suarez, Malte Ostendorff, Samuel Weinbach, Rafet Sifa, Stefan Kesselheim, Nicolas Flores-Herr
| Challenge: | Recent success of large language models has been driven by curating the training dataset composition, scaling of model architectures and advancements in pretraining objectives, leaving tokenizer influence as a blind spot. |
| Approach: | They conduct a comprehensive study on the influence of tokenizer choice on LLM downstream performance by training 24 mono- and multilingual LLMs at a 2.6B parameter scale. |
| Outcome: | The proposed model can significantly impact the model's downstream performance and training costs. |
Honey, I Shrunk the Language: Language Model Behavior at Reduced Scale. (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have focused on high-compute settings, leaving the question of when these abilities begin to emerge largely unanswered. |
| Approach: | They investigate whether effects of pre-training can be observed when problem size is reduced, modeling a smaller, reduced-vocabulary language. |
| Outcome: | The proposed model performance is correlated with pre-training perplexity and performance. |
Omni-RewardBench: Toward a Comprehensive Evaluation of Generative Reward Models Across Modalities (2026.acl-long)
Copied to clipboard
Chi-Min Chan, Yujin Zhou, Pengcheng Wen, Boqin Yin, Jiaming Ji, Juntao Dai, Wei Xue, Sirui Han, Yike Guo
| Challenge: | Existing evaluation benchmarks for ORMs are largely text-centric or limited to bimodal tasks . a new study examines the effectiveness of Omni-RewardBench for ORms across modalities . |
| Approach: | They propose a hybrid automatic-annotation and human-verification pipeline to construct high-quality evaluation data. |
| Outcome: | The proposed model is the first benchmark for comprehensive evaluation of ORMs across modalities. |
Pre-Training Curriculum for Multi-Token Prediction in Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Multi-token prediction (MTP) is a pre-training objective for language models . prior work has shown that smaller language models struggle with the MTP objective . |
| Approach: | They propose a curriculum learning strategy that uses multiple prediction heads to predict the next tokens at each prediction step. |
| Outcome: | The proposed curriculum improves performance and output quality while retaining the benefits of self-speculative decoding. |
Language Fusion for Parameter-Efficient Cross-lingual Transfer (2025.acl-long)
Copied to clipboard
| Challenge: | Limited availability of multilingual text corpora for pretraining results in poor performance on downstream tasks due to undertrained representation spaces for languages other than English. |
| Approach: | They propose a method that integrates source and target language representations within low-rank (LoRA) adapters using lightweight linear transformations to enhance representation quality and transfer performance for languages other than English. |
| Outcome: | The proposed method improves representation quality and performance for languages other than English while maintaining parameter efficiency. |