The paper introduces anchored factual recall, where language models must choose an object for a subject–relation pair conditioned on an explicit anchor such as a time. Using causal interventions on four open‑weight models, the authors identify two parallel computational routes: a subject‑mediated route that aligns the subject representation with the queried year, and a direct route that lets the final position attend to the anchor and uses upper‑layer MLPs to bias the answer. These routes combine to determine the model’s answer, showing that factual recall involves selecting among context‑dependent alternatives rather than retrieving a single stored object.
arXiv:2609.24238v1 Announce Type: new Abstract: We reproduce and stress-test the work of Yu et al. (2023), who characterize how language models (LMs) arbitrate between memorized knowledge and contrad...
The paper introduces “Rel-Bloom,” a multilingual diagnostic benchmark of 450 multiple-choice questions that tests language models’ ability to reason over relational facts supplied directly in prompts. It evaluates 43 models across six languages and two multilingual settings, revealing that accuracy drops sharply from Apply to Analyze levels, that Evaluate items remain a bottleneck even for high-performing models, and that difficulty rankings are consistent across languages. The study also finds that counterfactual contexts are generally harder than factual ones, and that removing entity identities can sometimes improve higher-level performance.
arXiv:2608. 15507v1 Announce Type: cross Abstract: A consistent concept of the current time is important for temporal reasoning, yet how language models represent the current time is not well understood.
arXiv:2607. 11327v1 Announce Type: cross Abstract: Model editing keeps large language models (LLMs) up to date without retraining, but temporal facts expose a limitation of the prevailing locate-and-edit paradigm: an update is not always a replacement.
arXiv:2609.25853v1 Announce Type: new Abstract: Learned-memory methods store information in an explicit table and consume it through a separate reader, allowing addressing, storage, and reading to be...
arXiv:2606. 01223v1 Announce Type: cross Abstract: Despite substantial progress in long-context modeling, existing benchmarks remain confined to factual memory for explicit recall, failing to measure the reflective memory required to synthesize fragmented, multimodal cues into high-level interpretations.
arXiv:2608. 14320v1 Announce Type: new Abstract: The anchoring effect is a cognitive bias in which an initial reference value shifts a later judgment toward itself.
The paper introduces Temporal State Tracking Memory (TEM), a training‑free framework that stores facts as versioned records linked to their sources and retrieves the relevant versions in source order, labeling historical records before passing them to the answer model. TEM incorporates event memory and original dialogue turns to fill gaps in structured facts. Evaluations on 420 questions from 600 sessions with 60 personas show high key‑entity hit rates (75.71–97.62%) and significant improvements when historical labels are added, while keeping evidence compact (94–118 tokens per answer).
arXiv:2610.27580v1 Announce Type: cross Abstract: Factual recall is a lookup: some representation inside the network serves as the key, and what that key is made of decides which queries can retrieve...
The paper introduces T‑Mem, a memory framework that uses temporal structure to manage long‑horizon multi‑party conversations. It employs Temporal Evidence Discovery to find locally concentrated evidence, Temporal Memory Structuring to model state transitions and compress low‑information spans, and Evidence Forest Linearization to serialize the evidence for answer generation. Experiments on two benchmarks show that T‑Mem outperforms existing memory frameworks across various settings.
arXiv:2608. 16168v1 Announce Type: cross Abstract: Large language model (LLM) agents increasingly use external memory systems to support personalization by drawing on long and evolving interaction histories, in which user preferences may be distributed across time, change with context, and conflict with earlier evidence.
The paper investigates how large language models (LLMs) incorrectly use post‑cutoff information when asked to answer from an earlier time, a problem termed temporal leakage. It shows that explicit cutoff statements and prefix constraints reduce leakage more effectively than implicit framing or suffix constraints, but that prompting alone cannot give models the ability to verify temporal admissibility. To overcome this, the authors introduce Temporal Critique Fine‑Tuning (TCFT), which trains models to detect and explain post‑cutoff leakage and judge whether a response is temporally admissible, achieving significant reductions in leakage on Qwen2.5 models.