Memory VLAs treat pauses as progress via temporal shortcuts
Brief pauses drop counting success from 47 to 16 of 50 and from 44 to 10; event-indexed memory recovers half or more without retraining.
Stories per day · 30 days
9 RSS
Relatedbenchmark4retrieval4inference3reinforcement-learning3agents2hidden-profile2huggingface2llm-agents2multi-agent2quantization2token-efficiency2video-understanding2agent1agent-benchmarks1agent-security1ai-agents1api1arxiv1
82 stories, newest first · one card per story
Stories tagged #memory
HiVe-Mem cuts multimodal agent memory tokens by about 90%
A graph of textual and visual anchors plus a trainable evidence selector improved three memory benchmarks across base models.
MaxIM frames incremental memory summarization as an RL reward-shaping problem
A variational mutual-information critic supplies dense rewards so bounded summaries stay predictive of future task targets.
Memorilla: Latent Semantic Memory for LLMs
arXiv paper introduces Memorilla for latent semantic memory in LLMs.
MemoryDojo benchmarks poisoning attacks on LLM agent memory
Benchmark with 155 trajectories demonstrates all tested memory systems are susceptible to poisoning.
LooM Enables Loop-Invariant Memory for Causal Encoder-Decoder Models
LooM loops only decoder layers after the first in causal encoder-decoder models to provide budgets from 3 to 21 executions per token without growing cache size.
LExGraph advances exploration memory
LExGraph organizes exploration dialogues into graphs for superior long-term memory.
VideoMemBench measures MLLM in-context video memory
VideoMemBench pairs videos with multi-turn dialogues containing episodic and answer dependencies to test memory recall in MLLMs.
DeepMem advances byte-level language modeling with multiscale architecture
DeepMem separates semantic compression from exact byte preservation using a learned invertible subband transform to create a hierarchy of shorter sequences.
One 1B decoder gates and writes memory-grounded conversation hints
End-to-end exact match is 0.2619 versus an oracle-gate ceiling of 0.2971 across 28,058 decision points.
Paper ties RL memory retrieval and credit to value of information
A learned world model estimates that quantity from archived rollouts, defining an optimal retrieval set and a TD-target floor.
Memory heuristics fail LLM group decisions on CrossProfile tasks
On 480 crossed hidden-profile tasks, relevance readmits decoys and rarest-first is no better than random on average.
Study tests how labels and memory change LLM agent envelope defenses
Authentic strong labels cut semantic breaches, but forged envelopes and prior success memory can erase that gain.
Memory heuristics fail LLM group decisions on CrossProfile
On 480 hidden-profile tasks, relevance and rarest-first memory selection both lose accuracy in predictable ways.
Test-time memory evolution beats test-time training on math discovery
TTME updates external memory instead of weights and cuts FLOPs by 82.8–84.6% and runtime by 47.7–58.4%.
PACE quantizes linear-attention recurrent states to four bits post hoc
Mixed four- and eight-bit PACE matches eight-bit DAMP on the three larger hybrid models at about two thirds of its memory.
SpatialBrain stores spatial-reasoning experience with a fixed backbone
On five benchmarks it beats direct answers from the same backbone on four and ties on the fifth.
Paper proposes memory-centric continual learning for foundation-model agents
MCCL judges memory by later task gains, with a taxonomy, two-loop lifecycle, eval protocol, and scaling model.
ChatGPT adds locked chats protected by PIN or biometrics
OpenAI says locking does not add end-to-end encryption, change retention, or clear prior Memory.
TokenWorth scores memory worth for lifelong agents under tight budgets
Across three LifelongAgentBench settings it records the highest success while emitting the fewest output tokens of the methods tested.
Warm-Start UCB personalizes memory gating for language agents
The policy learns which facts to keep from delayed retrieval feedback and beat baselines across memory budgets.
Peak-Memory Schedules You Can Trust: Exact Realization, Certified Optimality, and the Limits of Recomputation for Inference
Statement-level reordering certifies peak-memory-optimal inference schedules more reliably than regenerating compiler code.
Beyond Marginal Influence: Interaction-Aware Memory Selection for LLM Agents
JointCore finds jointly sufficient LLM-agent memories, recovering minimum-cost sets in all 6,912 exact benchmark cases.
Conditional Program stores LLM n-gram lookup tables at two bits
For DeepSeek, the method cuts table storage from 203GB to 51GB versus keeping full embeddings.