Mechanisms
Memory Systems · Declarative
Episodic Memory
Example
Someone told you last spring that they had started a new job. Remembering it brings back specifics: the coffee shop, that they were nervous, that you were surprised because you had not known they were looking. Years later you may still know where they work while retrieving none of that, or find yourself certain she told you something over coffee and unable to recover what. The knowing and the remembering come apart in both directions.
Episodic Memory in Humans
Definition
Episodic memory is a neurocognitive system that allows a person to re-experience a specific past event rather than merely know that it occurred (Tulving, 2002). Its defining property is autonoetic consciousness: awareness of oneself as having lived through the event, situated in subjective time. Semantic retrieval carries noetic awareness instead, which is knowing without re-experiencing (Tulving, 2002). Tulving is explicit that this is a system rather than a class of content or of task, and that it is systematically related to, but not identical with, either the information retrieved or the experience of retrieving it.
Neural Substrate
Ranganath and Ritchey (2012) propose two dissociable large-scale networks, converging across human, monkey, and rat work: an anterior temporal system centred on perirhinal cortex, which represents the entities in an event, and a posterior medial system centred on parahippocampal and retrosplenial cortex, which represents context. They propose, while noting that direct evidence is thin, that the hippocampus links the two. Their own summary table assigns episodic recollection to the posterior medial system and characterizes the anterior temporal system's role as semantic and familiarity-based, so this row cites them for the item and context division rather than for a localization of episodic memory as a whole.
The evidence for the item component itself converges across species. In monkeys, perirhinal lesions impair object recognition far more than hippocampal or parahippocampal lesions do; in rats, perirhinal immediate-early-gene expression tracks object familiarity but not spatial arrangement, with the reverse pattern in postrhinal cortex; in humans, perirhinal activation at encoding and retrieval tracks item familiarity and recollection of an item's own features, whereas parahippocampal activation tracks the context in which the item was studied (Ranganath & Ritchey, 2012). A single patient with damage largely confined to perirhinal and adjacent anterior temporal cortex, sparing hippocampus and parahippocampal cortex, showed impaired item familiarity with intact recollection of context, the dissociation the two-systems account predicts (Ranganath & Ritchey, 2012).
AI at a Glance
Humans hold memory for the specific occasions they lived through separately from memory for facts, so you can remember the afternoon someone told you about their new job, the coffee shop and what you thought at the time, without being able to say what the company is called (Tulving, 2002). Language model systems have no such separation by default: tested directly as what-where-when recall over a narrative they recover single events and fail once a cue matches several, and on conversational recall the best long-context and retrieval configurations trail the human score by roughly 36 points, with retrieval over extracted assertions edging out retrieval over the dialogue itself (Huet et al., 2025; Maharana et al., 2024).
Episodic Memory in AI
How Far It Has Been Measured
Measurement in AI: Directly Measured
AI Memory Performance: Below Human approximate provisional
Where It Lives in the System
What Has Been Measured
- A peer-reviewed temporal-order study (ICML 2026) recruited 97 Goodreads readers who had finished The Murder of Roger Ackroyd within the previous 30 days and, without the book, asked each which of two 50-word passages came first for about 10 pairs; people were near chance for nearby passages and reliably above chance for widely separated ones (Pink et al., 2026).
- Current conversation: Llama 3.1 8B and 70B (2024) judged the same 296 passage pairs with the whole novel in their context window, meaning they could still see the book; both reached at least 70% for far-apart passages, the 8B model never differed significantly from people (all p ≥ .289), and the 70B model beat people only for passages 16,000–24,000 words apart (p = .021), so under the uneven-test rule this counts as a tie and the one above-human bin is not counted (Pink et al., 2026).
- Training, the fair condition where neither side could look back: a 2024 preprint by the same group found Llama3-8B and Mistral-7B trained on the books for five epochs no better than copies trained without them (about 51%, near chance), while 155 readers averaged 64% on 50-word passages (56% on 20-word ones), were significantly above chance beyond 4,000 words, and peaked at 76% beyond 25,000 words; this condition was dropped from the peer-reviewed version (Pink et al., 2024).
- Saved memory and spread across systems in the same preprint: four models using retrieval (RAG, meaning the system searches stored passages and pastes matches into its input) scored 55–67% at distances up to 16,500 words, overlapping the human scores with no statistical comparison, while 11 models given book excerpts in view ranged from 56% (Llama2-7B) to 95% (Mixtral-8x22B), a range that does not count against people because the text was visible (Pink et al., 2024).
- The rating is Below Human, approximate and provisional: the two results disagree, and the training test, where neither side could look back at the book, better captures episodic memory, which is remembering an experience rather than consulting a record of it, so it outranks the tie made with the book in view; that test is the weaker study (humans read one book, models nine; only two small models), and the rating rests on one research group's task and 2023–2024 open models (Pink et al., 2026).
- A 2026 preprint gave 50 Prolific adults and nine models (including GPT-5.4 and Claude Opus 4.6) a short story with questions about event order and a naturalistic story free-recall task; people averaged 7.96 of 10 and a recall similarity of 0.60, models 9.22–9.80 and 0.68–1.00, but the story was hidden from people and stayed in view for the models, so these higher scores are not counted, and a memory-limited agent built to imitate people is a simulation rather than a test of ability (Wang et al., 2026).
- A peer-reviewed benchmark with no human baseline asked models to recall events by time, place, entity, and content from a synthetic book of about 100,000 tokens; with one matching event F1 (a combined accuracy score from 0 to 1) ranged from 0.05 (o1-mini) to 0.81 (GPT-4o) with the book in view, every model fell to 0.60 or less once two or more events matched the cue, and no model recalled the complete set of events for more than 18% of multi-event questions (Huet et al., 2025).
- Saved memory in commercial assistants: across short three-to-six-session chat histories, ChatGPT's memory feature answered 57.7% (GPT-4o) and 71.1% (GPT-4o-mini) of 97 questions correctly and Coze 33.0% and 24.7%, versus 91.8% when GPT-4o read the full history directly; the benchmark reports no human baseline (Wu et al., 2025).
- A peer-reviewed NeurIPS 2024 study found that attention heads in pretrained models favour the item that followed a repeated item, a temporal contiguity and forward bias resembling human free-recall curves (PEERS, N = 171); this is a mechanistic parallel to the temporal context model, not a comparison of how much either side remembers (Ji-An et al., 2024).
- A 2025 preprint gave GPT-4 (with Mistral and Llama checks) word lists from the Deese-Roediger-McDermott paradigm and found perfect recall with no primacy or recency effect, set against published human serial-position curves; because the list stayed in view for the model, exceeding people is not counted (Cao et al., 2025).
Where a Mechanism Can Live
04 Product
The whole configuration, and the only thing a benchmark observes directly.
01
Weights
Across deployments
02
Context Window
Within a session
03
Memory Layer
Across sessions
Which of the three a result belongs to is inferred from it, not seen in it.
Read more
Anterior temporal lobe
Lobe in the temporal lobe
The forward portion of the temporal lobe, converging on the temporal pole and heavily connected to orbitofrontal and limbic cortex.
The anterior temporal lobe covers the front of the superior, middle and inferior temporal gyri together with the temporal pole and the anterior fusiform gyrus. It sits at the end of the ventral visual and auditory streams and connects strongly to orbitofrontal cortex, amygdala and entorhinal cortex through the uncinate fasciculus. Its degeneration in semantic dementia produces progressive loss of word and object meaning across modalities, which motivated the "semantic hub" account. Whether the region is genuinely modality-general or an assembly of separable category-specific areas remains contested, partly because MRI signal here is badly distorted.
Hippocampus
Cortex in the hippocampal formation
A curved allocortical structure in the medial temporal lobe whose bilateral damage reliably produces dense anterograde amnesia in humans.
A paired, seahorse-shaped fold of three-layered cortex running front to back inside the temporal lobe, connected to the rest of the brain mainly through entorhinal cortex and the fornix. What it does is genuinely contested and has been for fifty years: the human lesion literature reads it as necessary for forming new declarative memories, the rodent recording literature as a map of space, and a third position treats both as instances of binding arbitrary elements into relations. The accounts differ on which species and which task they explain best, and the question of which is right remains open.
Parahippocampal cortex
Cortex in the parahippocampal gyrus
The posterior part of the parahippocampal gyrus, responding strongly to scenes and spatial layouts and feeding medial entorhinal cortex.
Cortex behind the entorhinal region on the medial temporal surface, connected to retrosplenial and posterior parietal areas and projecting into medial entorhinal cortex. Human imaging finds a patch here that responds far more to images of places than of objects or faces: the parahippocampal place area. What that response encodes is unsettled: candidate accounts include scene geometry, spatial context for navigation, and the presence of large landmark-like objects, and each explains some of the contrast that named the region.
Perirhinal cortex
Cortex in the medial temporal lobe
Cortex along the rhinal sulcus that relays object information to the hippocampus and is itself implicated in item familiarity.
Areas 35 and 36, flanking the rhinal sulcus and receiving heavy input from ventral visual stream regions carrying object identity. It projects onward to lateral entorhinal cortex and so into the hippocampus. The contested question is whether it is only a relay: one account holds that perirhinal cortex supports familiarity judgments independently of hippocampal recollection, which would split recognition memory into two processes with two substrates; another treats the dissociation as an artifact of memory strength. Both camps have patient data.
Retrosplenial cortex
Cortex in the cingulate cortex
A small strip of cortex tucked behind and below the splenium of the corpus callosum, Brodmann areas 29 and 30.
The retrosplenial cortex lies immediately behind the splenium (the thickened rear end of the corpus callosum), buried in the callosal sulcus, and is substantially smaller in humans than its prominence in rodent work suggests. It connects the hippocampal formation and anterior thalamus with parietal and visual cortex, making it a waypoint between memory and spatial systems. Damage produces topographical disorientation in which landmarks are recognized but cannot be used for heading. Rodent and human literatures use the term for somewhat different tissue, since the human region is displaced by the expansion of surrounding cortex.
Cao et al. (2025)
Analyzing Memory Effects in Large Language Models through the lens of Cognitive Psychology
In
arXiv
Abstract
Memory, a fundamental component of human cognition, exhibits adaptive yet fallible characteristics as illustrated by Schacter's memory "sins".These cognitive phenomena have been studied extensively in psychology and neuroscience, but the extent to which artificial systems, specifically Large Language Models (LLMs), emulate these cognitive phenomena remains underexplored. This study uses human memory research as a lens for understanding LLMs and systematically investigates human memory effects in state-of-the-art LLMs using paradigms drawn from psychological research. We evaluate seven key memory phenomena, comparing human behavior to LLM performance. Both people and models remember less...
Find it
Huet et al. (2025)
Episodic Memories Generation and Evaluation Benchmark for Large Language Models
In
arXiv
Abstract
Episodic memory -- the ability to recall specific events grounded in time and space -- is a cornerstone of human cognition, enabling not only coherent storytelling, but also planning and decision-making. Despite their remarkable capabilities, Large Language Models (LLMs) lack a robust mechanism for episodic memory: we argue that integrating episodic memory capabilities into LLM is essential for advancing AI towards human-like cognition, increasing their potential to reason consistently and ground their output in real-world episodic events, hence avoiding confabulations. To address this challenge, we introduce a comprehensive framework to model and evaluate LLM episodic memory...
Find it
Ji-An et al. (2024)
Linking In-context Learning in Transformers to Human Episodic Memory
In
arXiv
Abstract
Understanding connections between artificial and biological intelligent systems can reveal fundamental principles of general intelligence. While many artificial intelligence models have a neuroscience counterpart, such connections are largely missing in Transformer models and the self-attention mechanism. Here, we examine the relationship between interacting attention heads and human episodic memory. We focus on induction heads, which contribute to in-context learning in Transformer-based large language models (LLMs). We demonstrate that induction heads are behaviorally, functionally, and mechanistically similar to the contextual maintenance and retrieval (CMR) model of human episodic...
Find it
Maharana et al. (2024)
Evaluating Very Long-Term Conversational Memory of LLM Agents
In
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 13851-13870
Abstract
Paraphrase, not the publisher's wording. Prior work on long-term open-domain dialogue evaluates model responses over no more than five chat sessions, leaving the effectiveness of long-context models and retrieval augmentation in genuinely long-term conversation untested. The authors introduce a machine-human pipeline that generates high-quality very long-term dialogues by grounding LLM agent conversations in personas and temporal event graphs, with agents able to share and react to images. Human annotators verify and edit the output for long-range consistency and grounding. The resulting dataset, LoCoMo, supports an evaluation benchmark covering question answering, event summarisation,...
Find it
Pink et al. (2024)
Assessing Episodic Memory in LLMs with Sequence Order Recall Tasks
In
arXiv
Abstract
Current LLM benchmarks focus on evaluating models' memory of facts and semantic relations, primarily assessing semantic aspects of long-term memory. However, in humans, long-term memory also includes episodic memory, which links memories to their contexts, such as the time and place they occurred. The ability to contextualize memories is crucial for many cognitive tasks and everyday functions. This form of memory has not been evaluated in LLMs with existing benchmarks. To address the gap in evaluating memory in LLMs, we introduce Sequence Order Recall Tasks (SORT), which we adapt from tasks used to study episodic memory in cognitive psychology. SORT requires LLMs to recall the correct...
Find it
Ranganath & Ritchey (2012)
Two cortical systems for memory-guided behaviour
In
Nature Reviews Neuroscience, 13(10), 713-726
Abstract
Although the perirhinal cortex (PRC), parahippocampal cortex (PHC) and retrosplenial cortex (RSC) have an essential role in memory, the precise functions of these areas are poorly understood. Here, we review the anatomical and functional characteristics of these areas based on studies in humans, monkeys and rats. Our Review suggests that the PRC and PHC–RSC are core components of two separate large-scale cortical networks that are dissociable by neuroanatomy, susceptibility to disease and function. These networks not only support different types of memory but also appear to support different aspects of cognition.
Find it
Tulving (2002)
Episodic Memory: From Mind to Brain
In
Annual Review of Psychology, 53(1), 1-25
Abstract
Episodic memory is a neurocognitive (brain/mind) system, uniquely different from other memory systems, that enables human beings to remember past experiences. The notion of episodic memory was first proposed some 30 years ago. At that time it was defined in terms of materials and tasks. It was subsequently refined and elaborated in terms of ideas such as self, subjective time, and autonoetic consciousness. This chapter provides a brief history of the concept of episodic memory, describes how it has changed (indeed greatly changed) since its inception, considers criticisms of it, and then discusses supporting evidence provided by (a) neuropsychological studies of patterns of memory...
Find it
Wu et al. (2024)
LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory
In
arXiv
Abstract
Recent large language model (LLM)-driven chat assistant systems have integrated memory components to track user-assistant chat histories, enabling more accurate and personalized responses. However, their long-term memory capabilities in sustained interactions remain underexplored. We introduce LongMemEval, a comprehensive benchmark designed to evaluate five core long-term memory abilities of chat assistants: information extraction, multi-session reasoning, temporal reasoning, knowledge updates, and abstention. With 500 meticulously curated questions embedded within freely scalable user-assistant chat histories, LongMemEval presents a significant challenge to existing long-term memory...
Find it