Integrating external knowledge into Large Language Models (LLMs) iscrucial for many real-world applications, yet current methods like Retrieval-Augmented Generation (RAG) face limitations with broad, multi-source queries, while long-context models are computationally prohibitive.We introduce CacheNotes: Task-Aware Key-Value Cache Compression. Given a task description and a corpus, CacheNotes first generates a sequence of Compression-Planning-Tokens (CPTs), an offline task-focused distillation pass that identifies and organizes key information from the corpus. These CPTs are then used to guide a one-time compression of the corpus into a compact, reusable KV cache, which is then used alone at inference time to efficiently answer diverse, reasoning-intensive queries, eliminating repeated retrieval or context expansion.Experiments on LongBench show that, on Question-Answering tasks at a 20× compression, CacheNotes outperforms RAG by over 8 F1 points and reduces latency by over 4×. On RULER, it surpasses previous query-agnostic compression methods by 55 points, narrowing the gap to query-aware compression approaches. Additional results on real-world enterprise and synthetic datasets demonstrate its strong performance on multi-hop and broad-coverage queries.

Processing large tables provided in-context to LLMs is challenging due to token limits and information overload. While Retrieval-Augmented Generation can select relevant subsets externally, this work explores Key-Value (KV) cache compression as an alternative, applied directly to the linearized table during inference. We show that the LLM’s internal attention scores over the table context guides the retention of essential KV pairs, effectively compressing the processing context while preserving crucial relational information needed for complex queries. Experiments on Spider, WikitableQA, and QTSumm datasets validate the compression approach for in-context table processing, offering a promising path for improved table representation learning in LLMs.

Recent large language model applications, such as Retrieval-Augmented Generation and chatbots, have led to an increased need to process longer input contexts. However, this requirement is hampered by inherent limitations. Architecturally, models are constrained by a context window defined during training. Additionally, processing extensive texts requires substantial GPU memory. We propose a novel approach, Finch, to compress the input context by leveraging the pre-trained model weights of the self-attention. Given a prompt and a long text, Finch iteratively identifies the most relevant Key (K) and Value (V) pairs over chunks of the text conditioned on the prompt. Only such pairs are stored in the KV cache, which, within the space constrained by the context window, ultimately contains a compressed version of the long text. Our proposal enables models to consume large inputs even with high compression (up to 93x) while preserving semantic integrity without the need for fine-tuning.