15 papers in four lines of work. Each paper page has a short summary, a page-by-page reader and the PDF.
No paper matches that filter. Clear the box or pick another line.
How errors build up as a model generates, where they cluster, and how production systems correct for them outside the weights. Read in order: the three papers form one argument.
Treat every token as an equal, independent chance to fail and long outputs look doomed. Published measurements find about 9% of tokens depend on long-range context, and errors correlate, so reliability tracks key decisions, not output length.
No finite fix list covers every failure mode of open-ended LLM use. Inside one deployment, published taxonomies suggest failures recur in a small catalogue, so a sufficient fix library grows slowly with sequence length, then levels off.
Production LLM systems are increasingly fixed by editing prompts, rules, memories, skills and tools, not weights. Across about 130 surveyed systems, that loop runs mostly as patchwork, and none governs its fixes across organisations.
Rewriting prompts and documents into a compact symbolic form, and measuring which facts and relations survive the rewrite.
Rewriting text so each line holds one claim, with explicit symbols for cause and contrast, cut tokens by about two fifths and lost fewer answers than LLMLingua-2's token deletion, by the widest margin on fine details.
Rewriting retrieved passages as pipe-separated entity-relation clauses, entities kept verbatim, beat character deletion, truncation and random subsampling at the same token budget on three multi-hop benchmarks. At a fixed budget, the form of the kept text is not a detail.
A symbolic code's own checker accepted translations that changed the source. Scored against a constructed reference, trace translators kept less per byte than a compact structured format, and larger translators left the gap open.
Asked to rewrite passages into relational text at a set token budget, two LLM encoders landed in the registered band on 33 of 9,600 outputs, and their median output ran long. Compressors should be compared at delivered lengths.
Asking whether a ranking finds every statement an edit changes, not just most, reorders systems. Embedding reranking nudged BM25's recall up and completed fewer sets; one model scored well on recall while missing the edited sentence itself.
Reasoning traces that can be checked mechanically, and the measurements needed to tell whether training actually improves them.
If a model writes its math reasoning as tagged lines, a rule-based linter can recompute equations and check claims in SymPy instead of asking a model to reread them. It caught 195 of 196 injected errors; self-verification, about two thirds.
On a small model's math reasoning, traces a deterministic checker accepted had the right answer about one time in three, and training the model to write checkable traces raised acceptance while lowering accuracy.
Under GRPO's per-group standardization, a process term added to the outcome reward cannot be tuned: its weight cancels where a group's outcomes agree and is swamped where they differ. Endpoint gains need replication and a shuffle control.
Labelling, categorising and normalising customer conversations with LLMs under real budgets, and auditing the labels without a gold set.
An LLM labeler agrees more with silver labels written by a sibling model than with labels from other families. The contrast needs no human gold, so same-family consensus can be audited, and discounted, where gold is scarce.
A confidence gate lets an encoder answer roughly 87% of customer calls; with a five-label shortlist for the rest, cost falls more than 90% below an LLM reading the full list on every call. The shortlist needs its own prompt.
Recursively over-splitting each call category, then merging only the pairs an LLM confirms are duplicates, surfaced 234 subcategories against 111 from flat clustering. Redundancy was cheap, 10 merges among 244 proposals; coverage was not.
Rewriting each call once into short, attributed, tagged statements improved a supervised encoder without selection, and tag-based selection helped several weaker prompted readers further. With a distilled 0.6B normalizer, no large model sits in the serving path.
