DistillSpec enhances speculative decoding speed by 10-45% using knowledge distillation.

Yongchao Zhou, Kaifeng Lyu, Ankit Singh Rawat et al.

LLMBar evaluates LLM evaluators' instruction-following ability, revealing performance gaps.

Zhiyuan Zeng, Jiatong Yu, Tianyu Gao et al.

This paper investigates linear decompositions of Transformer embeddings, showing high correlation with model performance at corpus level but high variability across runs and tasks.

Timothee Mickus, Raúl Vázquez

Mistral 7B employs GQA and SWA, with 7B parameters, outperforming Llama 2 13B in reasoning and efficiency.

Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch et al.

SWE-bench evaluates LMs on real GitHub issues; Claude 2 solves only 1.96%.

Carlos E. Jimenez, John Yang, Alexander Wettig et al.

Proposes Sheared-LLaMA, a structured pruning method that compresses a 7B model to 1.3B and 2.7B parameters, achieving performance surpassing comparable models with only 3% training cost.

Mengzhou Xia, Tianyu Gao, Zhiyuan Zeng et al.

Selective Context method enhances inference efficiency by compressing context, reducing context cost by 50%.

Yucheng Li, Bo Dong, Chenghua Lin et al.

LLMLingua employs a hierarchical prompt compression framework, achieving up to 20x compression while maintaining near-original task performance.

Huiqiang Jiang, Qianhui Wu, Chin-Yew Lin et al.

FREE framework significantly reduces inference latency with synchronized parallel decoding, enhancing performance.

Sangmin Bae, Jongwoo Ko, Hwanjun Song et al.

DecoderLens introduces layerwise interpretation by cross-attending intermediate encoder states in encoder-decoder Transformers.

Anna Langedijk, Hosein Mohebbi, Gabriele Sarti et al.

PPNL benchmark shows GPT-4's potential in spatial reasoning but struggles with long-term reasoning.

Mohamed Aghzal, Erion Plaku, Ziyu Yao

Self-Taught Optimizer (STOP) recursively improves code generation, significantly enhancing performance.

Eric Zelikman, Eliana Lorch, Lester Mackey et al.

Training language models with pause tokens improves SQuAD task accuracy by 18%.

Sachin Goyal, Ziwei Ji, Ankit Singh Rawat et al.

OceanGPT, based on LLaMA-2, uses multi-agent collaboration for ocean science tasks, outperforming baselines in OceanBench evaluations.

Zhen Bi, Ningyu Zhang, Yida Xue et al.

LLMs fail to self-correct reasoning without external feedback, often degrading performance.

Jie Huang, Xinyun Chen, Swaroop Mishra et al.

PECORE framework quantifies context reliance plausibility in NMT via contrastive metrics and attribution, identifying context-sensitive tokens and their cues.

Gabriele Sarti, Grzegorz Chrupała, Malvina Nissim et al.

RoG framework integrates LLMs with knowledge graphs via a planning-retrieval-reasoning pipeline, achieving state-of-the-art KGQA performance with 85.7% Hits@1 on WebQSP.

Linhao Luo, Yuan-Fang Li, Gholamreza Haffari et al.

DAEMON formulates decoding as a multi-metric energy optimization, guaranteeing perplexity improvement and better alignment with human texts.

Haozhe Ji, Pei Ke, Hongning Wang et al.

RoleLLM enhances role-playing in LLMs via a four-stage framework, creating the RoleBench dataset.

Zekun Moore Wang, Zhongyuan Peng, Haoran Que et al.

Introduces COBBLER benchmark to evaluate biases in 16 large language models, revealing prevalent cognitive biases affecting evaluation robustness.

Ryan Koo, Minhwa Lee, Vipul Raheja et al.