datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
MemoryAgentBench
🚧 Update
(Sep 29th, 2025) We updated our paper, where we removed some in-efficient and high-cost samples. We also added a sub-sample of DetectiveQA.
(July 7th, 2025) We released the initial version of our datasets.
(July 22nd, 2025) We modify the datasets slightly, adding the keypoints in LRU and change the uuid into qa_pair_ids. The question_ids is only used in Longmemeval task.
(July 26th, 2025) We fixed bug on qa_pair_ids.
(Aug.5th, 2025) We removed the… See the full description on the dataset page: https://huggingface.co/datasets/ai-hyz/MemoryAgentBench.lldms-associative-memory-samples
LLDMs Associative Memory — Generated Samples
Model-generated text for the paper:
Language Diffusion Models are Associative Memories Capable of Retrieving Unseen Data
Bao Pham, Mohammed J. Zaki, Luca Ambrogioni, Dmitry Krotov, Matteo Negri
Accepted to EMNLP 2026 (Main Conference).
arXiv:2604.26841 · paper · code · checkpoints
29.5 million generated sequences (~3.8B tokens) sampled from the released checkpoints — one
generation run per (model size, training-set fraction). These… See the full description on the dataset page: https://huggingface.co/datasets/lemoncmd/lldms-associative-memory-samples.MemoryDecoder-at-Scale-domain-data
MemoryDecoder at Scale Domain Data
This repository contains the domain-specific continued-pretraining (CPT) data,
the tokenized and preprocessed datasets, and the aligned KNN distributions used
by MemoryDecoder at Scale.
Links
Project Page: Memory Decoder at Scale
GitHub Repository: LUMIA-Group/MemoryDecoder-at-Scale
Paper: Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory
The preprocessed datasets and KNN distributions in this repository use… See the full description on the dataset page: https://huggingface.co/datasets/Rubin-Wei/MemoryDecoder-at-Scale-domain-data.CoMEM-agent-memory-trajectories
📊 Auto-Scaling GUI Memory Dataset
This dataset accompanies our paper:📄 Auto-Scaling Continuous Memory for GUI Agent
We present a large-scale, diverse dataset for training and evaluating GUI-based agents with auto-scaling continuous memory. The dataset includes expanded web links, generated tasks, and executed trajectories, spanning a wide array of real-world domains.
📁 Dataset Structure
The dataset includes the following components:
✅ Expanded Links… See the full description on the dataset page: https://huggingface.co/datasets/WenyiWU0111/CoMEM-agent-memory-trajectories.MemoryAgentBench
🚧 Update
(Sep 29th, 2025) We updated our paper, where we removed some in-efficient and high-cost samples. We also added a sub-sample of DetectiveQA.
(July 7th, 2025) We released the initial version of our datasets.
(July 22nd, 2025) We modify the datasets slightly, adding the keypoints in LRU and change the uuid into qa_pair_ids. The question_ids is only used in Longmemeval task.
(July 26th, 2025) We fixed bug on qa_pair_ids.
(Aug.5th, 2025) We removed the… See the full description on the dataset page: https://huggingface.co/datasets/Robin076/MemoryAgentBench.agent-memory-bench-corpus
agent-memory-bench: the experience corpus
The neutral feed for a preregistered, execution-graded benchmark of memory layers for coding
agents. Every memory product under test ingests these same bytes through its own write path,
then an agent is given real coding work in a real repository where success depends on something
established in an earlier session, and the artifact is graded by execution: the task's tests
pass or they do not.
There is no LLM judge anywhere in the primary… See the full description on the dataset page: https://huggingface.co/datasets/Gde05/agent-memory-bench-corpus.tool-reasoning-sft-MEMORY-mem_agent-sft-data-cleaned-rectified-408k
mem_agent-sft-data-cleaned-rectified
Multi-turn long-context memory-agent SFT dataset with explicit reasoning traces, structured tool calls, and sequential chunk-processing sub-chains.
Schema
Column
Type
Description
messages
string (JSON)
JSON-serialized list of {role, content} dicts. Roles: system, user, reasoning, tool_call, tool_output, answer
core_chain_OR_subcall
string
"core_chain" (full orchestration trace) or "subcall" (single chunk-processing step)… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/tool-reasoning-sft-MEMORY-mem_agent-sft-data-cleaned-rectified-408k.brain-memory
🧠 NIFTY AI Agent: Memory OS Cloud Snapshot
Cloud backup repository for the NIFTY 50 Autonomous AI Agent Memory OS.
• Repository: nagarhimanshu37/brain-memory• Total Stored Records: 190• Last Synchronized: 2026-09-22 08:24:45 UTC
📊 Partition Statistics
Partition
Records
Description
conversation_memory
66
Multi-turn trader dialogues & intent logs
episodic_memory
30
Trading day episodes (facts vs interpretations)
experience_memory
50
Crystallized… See the full description on the dataset page: https://huggingface.co/datasets/nagarhimanshu37/brain-memory.MemoryCD
MemoryCD
Filtered cross-domain subset of Amazon Reviews 2023 for memory-augmented LLM
evaluation. All configs expose a single test split (evaluation only).
Contents
Config
Records
users_interactions
323 users
meta_personal_care
33,475 items
meta_books
48,054 items
meta_electronics
25,441 items
meta_home
60,900 items
The 4 meta files contain exactly the items referenced by the 323 users
(167,870 unique parent_asin, 100% coverage). The price field is… See the full description on the dataset page: https://huggingface.co/datasets/WZDavid/MemoryCD.prompt-policy-memory-v0
Prompt Policy Memory v0
Synthetic profile-memory data: 100 training sessions from10users;20test sessions from2fresh users. Test users were generated after the GRPO checkpoint was frozen and must not be used for training or tuning.
Each row includes cumulative plain-text session input, a canonical plain-text key:value reference, chat messages, and evaluator-only target data. messages can be used for supervised fine-tuning. The reference contains all currently revealed facts; it… See the full description on the dataset page: https://huggingface.co/datasets/flavianv/prompt-policy-memory-v0.omnimcp_sliding_window_memory_pruner_teaser
🔬 INSPECT THE DEEPSEEK-R1 REASONING CHAIN LIVE:
Zero hallucinations. Null syntax errors. 100% AST compiler validated.🌐 Live Interactive Reasoning & Code Inspector: https://emgena.com/trainingslager🎁 Claim your Free Starter Kit (Code: STARTER100): https://emgena.com/trainingslager🏷️ Launch Discount: Get 20 € OFF any 500-incident production suite with code LAUNCH20!
📜 Enterprise Compliance: EU AI Act Articles 50 & 53 certified • 100% DSGVO / GDPR clean • Commercial EULA… See the full description on the dataset page: https://huggingface.co/datasets/emgena/omnimcp_sliding_window_memory_pruner_teaser.recursive-memory-perfectblend-coding
Frozen PerfectBlend + Coding training data
Private migration snapshot of the raw, Qwen-generated training trajectories used
by the all-turn residual Compressor dataset frozen on 2026-09-06. This is not
the untouched upstream PerfectBlend dataset or a newly generated corpus.
Corpus
Trajectories
Generated assistant responses
Raw bytes
PerfectBlend / xhigh
40,596
51,039
549,854,699
Coding
37,560
86,880
1,274,359,351
Total
78,156
137,919
1,824,214,050
The mixture… See the full description on the dataset page: https://huggingface.co/datasets/mocoV3/recursive-memory-perfectblend-coding.MemoryDecoder-domain-data
Dataset Description
This dataset contains the test splits used to evaluate the Memory Decoder model across three specialized domains: biomedical, legal, and finance.
The test data was randomly sampled from publicly available datasets to assess the model's performance in domain-specific language understanding.
GitHub: https://github.com/LUMIA-Group/MemoryDecoder
Dataset Sources
The test data is sampled randomly from the following source datasets:… See the full description on the dataset page: https://huggingface.co/datasets/Clover-Hill/MemoryDecoder-domain-data.clinical-cross-modal-memory-fidelity-v0.1Clinical Cross-Modal Memory Fidelity v0.1
Goal
Test whether prior image evidence is recalled accurately over time
Detect retroactive distortion driven by later narrative
Detect fabrication used to patch memory gaps
What it measures
memory_driftEarlier image facts are altered or inverted
fabricationNew findings are invented at recall
cross_modal_consistencyRecalled description matches original image evidence
How it works
Initial image facts are fixed and explicit
Intervening tasks… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-cross-modal-memory-fidelity-v0.1.swebench-verified-kimi-k2p6-traces
SWE-bench Verified Kimi K2.6 Reasoning Traces
This dataset contains reasoning traces generated on princeton-nlp/SWE-bench_Verified using fireworks_ai/kimi-k2p6-high with a mini-swe-agent based harness. It is intended for research and distillation of software-engineering agents.
The repository is published with three configs because each table has a different schema:
raw_trajectories: one row per SWE-bench instance with the patch, sanitized result JSON, full trajectory JSON, message… See the full description on the dataset page: https://huggingface.co/datasets/MemoryAsModality/swebench-verified-kimi-k2p6-traces.memory_sft_data
memory_sft_data
SFT data that teaches an agent to manage its own context window while solving
software-engineering tasks: gather the right code, compress aggressively with an
edit_context tool (offloading stale output to a memory store and leaving a short
self-contained note), and reuse that offloaded memory like a retrieval datastore
(ls/grep/cat over /tmp/.unified_memory/), then write a precise, grounded fix
plan that recalls offloaded details.
Each example is a full… See the full description on the dataset page: https://huggingface.co/datasets/rulins/memory_sft_data.memory-reasoning-split-eval-sets
Memory-Reasoning-Split Stage D Eval Sets
Curated + subsetted closed-book evaluation sets used to measure the per-domain
factual degradation × reasoning retention trade-off in the
memory_reasoning_split
Stage D/E forget-corpus ablation.
All factual rows share a unified schema so a single evaluator can score them:
{
"question": str,
"aliases": list[str], # any normalized alias match counts as a hit
"relation": str,
"topic": str,
"source_dataset": str (popqa_general /… See the full description on the dataset page: https://huggingface.co/datasets/hyunseoki/memory-reasoning-split-eval-sets.MemoryAgentBench
🚧 Update
(Sep 29th, 2025) We updated our paper, where we removed some in-efficient and high-cost samples. We also added a sub-sample of DetectiveQA.
(July 7th, 2025) We released the initial version of our datasets.
(July 22nd, 2025) We modify the datasets slightly, adding the keypoints in LRU and change the uuid into qa_pair_ids. The question_ids is only used in Longmemeval task.
(July 26th, 2025) We fixed bug on qa_pair_ids.
(Aug.5th, 2025) We removed the… See the full description on the dataset page: https://huggingface.co/datasets/Kakezh/MemoryAgentBench.Blum-Finance-Memory
BLUM Finance Memory
Opt-in community contributions for future BLUM Finance research.
Every contribution is quarantined. Uploading an example does not update the released
model, BLUM Engine memory, trading weights or production rules. Contributions must
pass schema, privacy, provenance, licensing, deduplication, poisoning and outcome
quality checks before they can enter a future versioned training dataset.
Installation and inference collect no telemetry. A contribution is created… See the full description on the dataset page: https://huggingface.co/datasets/Italianhype/Blum-Finance-Memory.