English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large Language Models (LLMs) prompted to generate chain-of-thought (CoT)exhibit impressive reasoning capabilities. Recent attempts at promptdecomposition toward solving complex, multi-step reasoning problems depend onthe ability of the LLM to simultaneously decompose and solve the problem. Asignificant disadvantage is that foundational LLMs are typically not availablefor fine-tuning, making adaptation computationally prohibitive. We believe (anddemonstrate) that problem decomposition and solution generation are distinctcapabilites, better addressed in separate modules, than by one monolithic LLM.We introduce DaSLaM, which uses a decomposition generator to decompose complexproblems into subproblems that require fewer reasoning steps. These subproblemsare answered by a solver. We use a relatively small (13B parameters) LM as thedecomposition generator, which we train using policy gradient optimization tointeract with a solver LM (regarded as black-box) and guide it throughsubproblems, thereby rendering our method solver-agnostic. Evaluation onmultiple different reasoning datasets reveal that with our method, a 175billion parameter LM (text-davinci-003) can produce competitive or even betterperformance, compared to its orders-of-magnitude larger successor, GPT-4.Additionally, we show that DaSLaM is not limited by the solver's capabilitiesas a function of scale; e.g., solver LMs with diverse sizes give significantperformance improvement with our solver-agnostic decomposition technique.Exhaustive ablation studies evince the superiority of our modular finetuningtechnique over exorbitantly large decomposer LLMs, based on prompting alone.

Small Language Models Fine-tuned to Coordinate Larger Language Models improve Complex Reasoning