{"arxiv_id":"2606.16285","incoming":{"papers":1,"sentences":2,"groups":[{"kind":"background","label":"Cites it as background","papers":1,"sentences":1,"quotes":[{"arxiv_id":"2607.26604","title":"WikiLoop: Jointly Learning to Build and Navigate Agent-Native Wikis with Downstream Feedback","url":"/paper/2607.26604","section":"Related Work","sentence":"HiMPO estimates the utility of an in-trajectory memory update through before/after recoverability under a shared pre-write state and assigns hindsight-filtered credit to the memory-writing tokens [ref] .","intent_p":0.83,"contests_p":null}]}],"unlabelled_sentences":1},"outgoing":{"papers":15,"sentences":21,"groups":[{"kind":"contests","label":"Contests","papers":2,"sentences":2,"quotes":[{"arxiv_id":"2402.03300","title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","url":"/paper/2402.03300","section":"Related Work","sentence":"Value-free RL methods such as GTPO [ref] and GRPO [ref] estimate advantages from group rewards, but often fail to identify pivotal intermediate decisions.","intent_p":0.42,"contests_p":0.76},{"arxiv_id":"2508.04349","title":null,"url":"/paper/2508.04349","section":"Related Work","sentence":"Value-free RL methods such as GTPO [ref] and GRPO [ref] estimate advantages from group rewards, but often fail to identify pivotal intermediate decisions.","intent_p":0.42,"contests_p":0.76}]},{"kind":"comparison","label":"Compares with","papers":7,"sentences":7,"quotes":[{"arxiv_id":"2502.12110","title":"A-MEM: Agentic Memory for LLM Agents","url":"/paper/2502.12110","section":"Experimental Setup","sentence":"Across the two protocols, we compare with ReAct [ref] , ReSearch [ref] , DeepResearcher [ref] , GRPO without memory [ref] , SFT-only memory agents, [ref] , SUPO [ref] , MEM1 [ref] , and A-MEM [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2402.03300","title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","url":"/paper/2402.03300","section":"Experimental Setup","sentence":"Across the two protocols, we compare with ReAct [ref] , ReSearch [ref] , DeepResearcher [ref] , GRPO without memory [ref] , SFT-only memory agents, [ref] , SUPO [ref] , MEM1 [ref] , and A-MEM [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2503.19470","title":"ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning","url":"/paper/2503.19470","section":"Experimental Setup","sentence":"Across the two protocols, we compare with ReAct [ref] , ReSearch [ref] , DeepResearcher [ref] , GRPO without memory [ref] , SFT-only memory agents, [ref] , SUPO [ref] , MEM1 [ref] , and A-MEM [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2504.03160","title":"DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments","url":"/paper/2504.03160","section":"Experimental Setup","sentence":"Across the two protocols, we compare with ReAct [ref] , ReSearch [ref] , DeepResearcher [ref] , GRPO without memory [ref] , SFT-only memory agents, [ref] , SUPO [ref] , MEM1 [ref] , and A-MEM [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2506.15841","title":"MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents","url":"/paper/2506.15841","section":"Experimental Setup","sentence":"Across the two protocols, we compare with ReAct [ref] , ReSearch [ref] , DeepResearcher [ref] , GRPO without memory [ref] , SFT-only memory agents, [ref] , SUPO [ref] , MEM1 [ref] , and A-MEM [ref] .","intent_p":1.0,"contests_p":null}]},{"kind":"background","label":"Cites as background","papers":7,"sentences":8,"quotes":[{"arxiv_id":"2506.15841","title":"MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents","url":"/paper/2506.15841","section":"Related Work","sentence":"SUPO [ref] incorporates summarization-based context management into multi-turn RL, MEM1 [ref] studies memory-reasoning synergy, and MemPO [ref] introduces a action optimized with trajectory- and memory-level advantages.","intent_p":0.96,"contests_p":null},{"arxiv_id":"2510.06727","title":null,"url":"/paper/2510.06727","section":"Related Work","sentence":"SUPO [ref] incorporates summarization-based context management into multi-turn RL, MEM1 [ref] studies memory-reasoning synergy, and MemPO [ref] introduces a action optimized with trajectory- and memory-level advantages.","intent_p":0.96,"contests_p":null},{"arxiv_id":"2603.00680","title":"MemPO: Self-Memory Policy Optimization for Long-Horizon Agents","url":"/paper/2603.00680","section":"Related Work","sentence":"SUPO [ref] incorporates summarization-based context management into multi-turn RL, MEM1 [ref] studies memory-reasoning synergy, and MemPO [ref] introduces a action optimized with trajectory- and memory-level advantages.","intent_p":0.96,"contests_p":null},{"arxiv_id":"2509.13313","title":null,"url":"/paper/2509.13313","section":"Related Work","sentence":"Summarization-based methods, such as ReSum [ref] , instead compress accumulated histories into compact reasoning states.","intent_p":0.9,"contests_p":null},{"arxiv_id":"2502.12110","title":"A-MEM: Agentic Memory for LLM Agents","url":"/paper/2502.12110","section":"Related Work","sentence":"Retrieval-based systems, such as MemGPT [ref] , MemoryBank [ref] , Mem0 [ref] and A-MEM [ref] , store past interactions in external memory and retrieve relevant fragments when needed.","intent_p":0.98,"contests_p":null}]}],"unlabelled_sentences":4},"claims":[],"thresholds":{"intent_p":0.8,"supports_p":0.95,"contests_p":0.7},"method":"Citing sentences come from each paper's LaTeX source; every \\cite marks the sentence and the works it cites. Pith's classifier typed each sentence. A label is shown at intent probability 0.8 or more (0.95 for support), a contest at 0.7 or more. Measured accuracy against two blind model families is at https://pith.science/accuracy. The quoted sentences are the citing authors' own.","held_up":null,"_links":{"html":"https://pith.science/paper/2606.16285"}}