The paper presents a rate‑distortion framework for understanding factual hallucination in closed‑book question answering. It shows that even when a fact is observed, limited memory forces it to be stored approximately, leading to errors that can be bounded by a combination of compression distortion and missing coverage. The authors derive a theoretical lower bound on error and validate it with simulations and probes on modern language models.
arXiv:2607. 18305v1 Announce Type: cross Abstract: Some limits on what language models know are not gaps in data coverage but structural properties of learning from text.
arXiv:2607. 04223v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) reduces but does not eliminate hallucination, and existing detectors return a single answer-level score that does not indicate which sentence is unsupported, or why.
arXiv:2607. 11020v1 Announce Type: cross Abstract: Continual learning promises a language model that keeps acquiring knowledge after training, with each new fact written into its weights.
arXiv:2605.10795v2 Announce Type: replace-cross Abstract: Large language models excel at factual recall, yet the fundamental limits of storing input--output associations with neural networks remain u...
arXiv:2607. 12649v1 Announce Type: new Abstract: Recent work on extractable memorization in LLMs suffers from two contrasting validity problems.
The paper presents a systematic study of how different compression techniques—pruning, quantization, and distillation—affect the capabilities of large language models (LLMs) in tasks such as mathematics, code generation, and question answering. It introduces a framework that measures capability loss and relates it to factors like model size, training stage, and compression settings, yielding simple predictive relations that generalize across unseen configurations. The authors demonstrate that sharing density responses across pruning levels can dramatically reduce the number of measurements needed, and that their predictive models closely match regression results while offering efficient decision guidance for compression method selection.
The paper introduces ResMem, a method that transforms retrieved knowledge into a parametric memory for large language models. By using the LLM as an offline teacher that reads retrieved text, ResMem learns an additive logit residual, avoiding large token storage and improving question‑answering performance. Experiments on Wikipedia and five QA datasets show significant storage savings and higher token F1 scores compared to k‑NN baselines.
The paper reports a Size‑Fidelity Paradox in large‑language‑model (LLM) context compression: larger compressors can produce lower reconstruction error yet yield less faithful reconstructed contexts. Two key mechanisms drive this effect: knowledge overwriting, where larger models replace source facts with their own beliefs, and semantic drift, where they paraphrase or restructure content. Across 27 compressor setups, mid‑sized models often recover contexts more faithfully, and analysis of embedding geometry shows that larger models organize memory into broader semantic subspaces, leading to ambiguous representations that weaken recovery.
arXiv:2608.29934v1 Announce Type: cross Abstract: KV-cache compression reduces LLM inference memory by evicting context tokens, but when the evicted tokens contain answer-bearing evidence, the model...
arXiv:2611.08341v1 Announce Type: new Abstract: AI agents store text as embedding vectors and compress them at scale. A compression is accepted when each item's nearest neighbors survive, but agents...
arXiv:2605. 05686v3 Announce Type: replace Abstract: Language models draw on two knowledge sources: facts baked into weights (parametric memory, PM) and information in context (working memory, WM).
The paper investigates why large language models often answer factual questions incorrectly in closed‑book settings. It shows that many errors are due to first‑token recall failures: the model commits to an incorrect first token but can complete the correct answer once the right token is provided. Using a First‑Token Cueing probe, the authors demonstrate that correcting just the first token recovers a significant portion of errors across 16 open models and improves accuracy on multilingual benchmarks.