The paper investigates how recursive contamination—retraining language models on their own generated text—affects output diversity across 13 publicly released checkpoints. Using a fixed contamination protocol over five generations, the authors find a wide spread in 4‑gram diversity (0.187 to 0.940), indicating that some models collapse into repetitive fragments while others remain largely unaffected. The study shows that a model’s susceptibility to collapse is an intrinsic property of the checkpoint, not predicted by parameter scale or static indicators, and that simple interventions such as tightening top‑p sampling can significantly slow or halt collapse.
By Yangze Liu, Zhongyi Han
The paper investigates neural text degeneration by measuring the fixed‑point structure of short‑window argmax maps across 17 pretrained models, using 96 random two‑token starts without prompts. It finds a stable four‑way classification that varies across model families and scales, with some models funneling to a single endpoint token while others do not, and shows that this behavior is not solely determined by training data or corpus frequency. The study demonstrates that repetition phenomena are not uniformly explained by either training data or network architecture alone, highlighting the complexity of neural text generation dynamics.
By Nicol\'as Vera Z\'u\~niga
arXiv:2607. 01480v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR), along with recent selfdistillation variants such as SDPO, evaluates each rollout against a verifier and updates the policy from that episode-level signal.
By Ye Liu, Srijan Bansal, Bo Pang, Yang Li, Zeyu Leo Liu, Yifei Ming, Zixuan Ke, Shafiq Joty, Semih Yavuz
arXiv:2509.15194v4 Announce Type: replace-cross
Abstract: Large language models (LLMs) can improve through reinforcement learning with verifiable rewards (RLVR), but many settings lack reference answ...
By Yujun Zhou, Zhenwen Liang, Haolin Liu, Wenhao Yu, Kishan Panaganti, Linfeng Song, Dian Yu, Xiangliang Zhang, Haitao Mi, Dong Yu
The paper investigates how preference optimization can inadvertently create a shared directional shift—termed a common‑mode offset—in a model’s hidden states. This offset, which affects both chosen and rejected responses, accumulates without being penalized by the ranking loss, eventually projecting onto the output layer to form a vocabulary well that dominates decoding and leads to generation collapse. Experiments across multiple base‑model lineages and fine‑tuning regimes confirm that masking this offset or using per‑token ranking losses can prevent the collapse, while standard ranking metrics fail to detect the risk.
By Yun Li, Ehsan Javanmardi, Simon Thompson, Jiachen Gong, Qunli Zhang, Zifan Zeng, Shiming Liu, Peng Wang, Zixuan Guo, Manabu Tsukada
arXiv:2610.06940v1 Announce Type: new
Abstract: Continual adaptation of language models can change their output distribution on prompts learned earlier, while retaining every old prompt-answer pair m...
By Huan Li, Zhe Cao, Qinlei Xie, Fushun Cui, Xuechen Liang
arXiv:2603. 11784v2 Announce Type: replace Abstract: As scaling laws push the training of frontier large language models (LLMs) toward ever-growing data requirements, training pipelines are approaching a regime where much of the publicly available online text may be consumed.
By Giorgio Racca, Michal Valko, Amartya Sanyal
The study measures the exact probability that language models assign to each integer from 1 to N when prompted for a random number, eliminating sampling noise by summing over all spellings. Across 18 open‑weight models and 29 decoding settings, instruction tuning consistently makes the distribution less uniform, introducing a round‑number avoidance heuristic and sharpening endpoint avoidance. While tuned models rarely output the extreme values 1 or N and largely avoid multiples of 10, they do not converge on a single favorite integer across models.
By Samar Ansari
The paper introduces a 400‑step ledger task that evaluates language models on three granular components: duplicate recognition, local processing, and counter tracking. Experiments on Qwen3‑8B show that duplicate‑recognition errors increase only late in the episode, while the counter remains flawless over 8,000 steps and local processing stays stable. Adding irrelevant content before each step causes the counter to fail consistently, but an external reminder can restore its accuracy.
By Thomas Serval
arXiv:2608. 19893v1 Announce Type: cross Abstract: Where does the novelty a base language model produces with no task come from, and what can an LLM judge of a long stream actually see?
By Roberto I. Ono Filho
Chronicle introduces a method called cut‑point replay to make regression testing of large language model (LLM) agents reproducible. It records an agent’s run at non‑deterministic boundaries as immutable envelopes and then replays selected boundaries while executing the rest live, enabling continuous‑integration tests that detect faulty code changes. Benchmarks show minimal overhead, perfect bit‑stability, and effective detection of unsafe actions in a mutation study.
By Tisha Chawla, Susheem Koul
arXiv:2606. 15621v1 Announce Type: new Abstract: Per-token counterfactual credit estimation asks which token in a language-model rollout caused the final answer to be right or wrong: cut the transcript at a pivot, substitute an alternative token, replay continuations, and compare outcomes.
By Nils Matteson