Escaping Mode Collapse in LLM Generation via Geometric Regulation
Mirrored from arXiv — NLP / Computation & Language for archival readability. Support the source by reading on the original site.
Computer Science > Computation and Language
Title:Escaping Mode Collapse in LLM Generation via Geometric Regulation
Abstract:Mode collapse is a persistent challenge in generative modeling and appears in autoregressive text generation as behaviors ranging from explicit looping to gradual loss of diversity and premature trajectory convergence. We take a dynamical-systems view and reinterpret mode collapse as reduced state-space accessibility caused by *geometric collapse*: during generation, the model's internal trajectory becomes confined to a low-dimensional region of its representation space. This implies mode collapse is not purely a token-level phenomenon and cannot be reliably solved by symbolic constraints or probability-only decoding heuristics. Guided by this perspective, we propose *Reinforced Mode Regulation* (RMR), a lightweight, online state-space intervention that regulates dominant self-reinforcing directions in the Transformer value cache (implemented as low-rank damping). Across multiple large language models, RMR substantially reduces mode collapse and enables stable generation at extremely low entropy rates (down to 0.8 nats/step), whereas standard decoding typically collapses near 2.0 nats/step.
| Comments: | Accepted to ICML 2026 |
| Subjects: | Computation and Language (cs.CL); Disordered Systems and Neural Networks (cond-mat.dis-nn); Artificial Intelligence (cs.AI); Chaotic Dynamics (nlin.CD) |
| Cite as: | arXiv:2605.00435 [cs.CL] |
| (or arXiv:2605.00435v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.00435
arXiv-issued DOI via DataCite
|
Submission history
From: Xin Du PhD [view email][v1] Fri, 1 May 2026 06:12:05 UTC (5,335 KB)
[v2] Wed, 27 May 2026 13:56:08 UTC (10,637 KB)
[v3] Fri, 31 Jul 2026 05:11:04 UTC (5,338 KB)
Access Paper:
- View PDF
- HTML (experimental)
- TeX Source
Current browse context:
References & Citations
Bookmark
Bibliographic and Citation Tools
Code, Data and Media Associated with this Article
Demos
Recommenders and Search Tools
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
More from arXiv — NLP / Computation & Language
-
Unified Hallucination Fuzzing for Multimodal Large Language Models
Aug 11
-
DocAtlas: Long-Document Understanding as Mutable-State Interaction
Aug 11
-
WuYuEval: A Multi-Level Benchmark for Large Language Models in Solid Waste Management
Aug 11
-
Search-G1: Grounded Search Agents via Representation-Based Intrinsic Rewards
Aug 11
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.