Public story · 2026-03-03 · source-backed

Memory-R1: RL-Based Agent Memory Management

Issue

2026-03-03

Confidence

source-backed

Sources

Redaction

passed

Story

Two specialized RL agents — Memory Manager (ADD/UPDATE/DELETE operations) and Answer Agent — fine-tuned with PPO and GRPO. With only 152 training QA pairs, outperforms baselines across three benchmarks. Directly applicable to persistent agent memory systems. arXiv 2508.19828

EMPO2: Memory-Augmented Agent (ICLR 2026)
Hybrid on/off-policy RL giving agents non-parametric memory for exploration. 128.6% improvement over GRPO on ScienceWorld, 11.3% on WebShop. Agents generalize to out-of-distribution tasks with "only a few trials with memory and no parameter updates." (arXiv)
Unsloth makes single-H100 long-context RL fine-tuning actually tractable.
New batching algorithms enable ~7x, up to 12x+, longer-context GRPO training with no accuracy or speed penalty versus optimized FA3 and chunked-loss setups (Unsloth Docs). Qwen3-8B GRPO reaches 110K context on one 80GB H100 via vLLM plus QLoRA. For solo builders doing reasonin...
A 2B-Parameter Model Learns When to Stop Answering
arXiv 2607.28457 is an oracle-free multi-turn RL framework where the model emits a solution plus a discrete correctness verdict and a confidence score each turn, keeping its answer only when the verdict is Correct and confidence clears threshold. Ground-truth correctness shape...
CogEvol reports a caught reward-hacking episode alongside its production numbers.
It turns a course brief into finished slides or a self-contained interactive HTML page in one pass (arXiv 2608.30968). Across 220k production requests the median slide takes 17 seconds and an interactive page 59. A hybrid rule-plus-VLM reward drives GRPO, hardened after the te...
Dark Souls Learning Environment: five methods, 22 bosses, every RL agent scores essentially zero.
arXiv 2608.09902 wraps all 22 boss encounters of Dark Souls: Remastered in a containerized Gymnasium-style benchmark where each step is a real action against the running game. On DSLE-5, an expert system and an evolutionary baseline beat only the tutorial boss (63% and 43% pea...
Gradient open-sources a full RL training loop for research agents
RobertGolds1/Gradient, created August 23 and at 423 stars in about a day, Apache-2.0, built on OpenPipe ART. It ships a Research Environment containing a reproducible company workspace of emails, contracts, policies, meeting notes and customer records with evidence deliberatel...

Source trail

arXiv 2508.19828 More from arxiv.org

Entities

Answer Agent Based Agent Memory Management Two Directly GRPO Memory Memory Manager PPO UPDATE

Provenance

Canonical issue: Ramsay Research Agent — 2026-03-03
AI generated: no
Story unit: 2026-03-03-memory-r1-rl-based-agent-memory-management
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

Memory-R1: RL-Based Agent Memory Management

Story

Related stories

EMPO2: Memory-Augmented Agent (ICLR 2026)

Unsloth makes single-H100 long-context RL fine-tuning actually tractable.

A 2B-Parameter Model Learns When to Stop Answering

CogEvol reports a caught reward-hacking episode alongside its production numbers.

Dark Souls Learning Environment: five methods, 22 bosses, every RL agent scores essentially zero.

Gradient open-sources a full RL training loop for research agents

Source trail

Entities

Provenance

Memory-R1: RL-Based Agent Memory Management

Story

Related stories

EMPO2: Memory-Augmented Agent (ICLR 2026)

Unsloth makes single-H100 long-context RL fine-tuning actually tractable.

A 2B-Parameter Model Learns When to Stop Answering

CogEvol reports a caught reward-hacking episode alongside its production numbers.

Dark Souls Learning Environment: five methods, 22 bosses, every RL agent scores essentially zero.

Gradient open-sources a full RL training loop for research agents

Source trail

Entities

Provenance

Story

Related stories

EMPO2: Memory-Augmented Agent (ICLR 2026)

Unsloth makes single-H100 long-context RL fine-tuning actually tractable.

A 2B-Parameter Model Learns When to Stop Answering

CogEvol reports a caught reward-hacking episode alongside its production numbers.

Dark Souls Learning Environment: five methods, 22 bosses, every RL agent scores essentially zero.

Gradient open-sources a full RL training loop for research agents

Tomorrow's issue, in your inbox.

Source trail

Entities

Provenance

Story

Related stories

EMPO2: Memory-Augmented Agent (ICLR 2026)

Unsloth makes single-H100 long-context RL fine-tuning actually tractable.

A 2B-Parameter Model Learns When to Stop Answering

CogEvol reports a caught reward-hacking episode alongside its production numbers.

Dark Souls Learning Environment: five methods, 22 bosses, every RL agent scores essentially zero.

Gradient open-sources a full RL training loop for research agents

Tomorrow's issue, in your inbox.

Source trail

Entities

Provenance