{"slug": "heavy-tailed-memory-traces-in-long-horizon-language-agents", "title": "Heavy-Tailed Memory Traces in Long-Horizon Language Agents", "summary": "A new arXiv paper (2610.00010v1) proposes Core-Tail World Model (CTWM), a rank-based memory controller that allocates prompt budget with a single exponent τ while retaining a summarized tail, after a tail audit found that semantic LLM policies yield the strongest truncated-power-law-compatible core-tail memory traces. On Synthetic Graph World, CTWM preserved full state and transition coverage, cut prompt tokens by 5.9%, and lowered bottom-half tail prediction error by 13.6% versus a graph-memory baseline, with a 24.48% token reduction on LongMemEval at aggregate accuracy parity. The authors argue heavy-tailed memory traces are both a diagnostic of finite retrieval and a practical control signal for token-efficient long-horizon language agent world models.", "body_md": "arXiv:2610.00010v1 Announce Type: new \nAbstract: Long-horizon language agents increasingly rely on external memory as a frozen world model, yet current memory systems are usually judged only by task success or token cost. We argue that the missing object is the shape of memory use: under finite context and repeated retrieval, agent memory can concentrate on a small core while leaving rare states in a long tail where prediction errors accumulate. We study this effect through a conservative tail audit and find that concentration is reproducible but policy-dependent. Random-walk agents produce log-normal-compatible retrieval artifacts, whereas semantic LLM policies yield the strongest truncated-power-law-compatible core--tail traces. Motivated by this audit, we propose Core--Tail World Model (CTWM), a rank-based memory controller that allocates prompt budget with a single exponent $\\tau$ while retaining a summarized tail. On Synthetic Graph World, CTWM preserves full state and transition coverage, reduces prompt tokens by 5.9%, and lowers bottom-half tail prediction error by 13.6% relative to a graph-memory baseline. The same paired comparison gives consistent token savings on ALFWorld and a 24.48% token reduction on LongMemEval with aggregate accuracy parity. These results suggest that heavy-tailed memory traces are not only a diagnostic of finite retrieval, but also a practical control signal for token-efficient agent world models.", "url": "https://wpnews.pro/news/heavy-tailed-memory-traces-in-long-horizon-language-agents", "canonical_source": "https://arxiv.org/abs/2610.00010", "published_at": "2026-10-02 04:00:00+00:00", "updated_at": "2026-10-02 04:14:56.504472+00:00", "lang": "en", "topics": ["ai-agents", "large-language-models", "ai-research", "artificial-intelligence"], "entities": ["Core-Tail World Model", "CTWM", "arXiv", "Synthetic Graph World", "ALFWorld", "LongMemEval"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/heavy-tailed-memory-traces-in-long-horizon-language-agents", "markdown": "https://wpnews.pro/news/heavy-tailed-memory-traces-in-long-horizon-language-agents.md", "text": "https://wpnews.pro/news/heavy-tailed-memory-traces-in-long-horizon-language-agents.txt", "jsonld": "https://wpnews.pro/news/heavy-tailed-memory-traces-in-long-horizon-language-agents.jsonld"}}