Pith. sign in

REVIEW 2 major objections 3 minor 167 cited by

LaMDA: Language Models for Dialog Applications

T0 review · 2 major / 3 minor · reviewed 2026-05-12 · grok-4.3

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fine-tuning LaMDA models on annotated human values plus access to external tools markedly raises safety and factual grounding in dialog responses.

desk verdict LaMDA shows that fine-tuning on safety annotations plus tool access improves dialog safety and grounding, but only on a narrow illustrative set of values and a basic trio of tools. read the letter →

arxiv 2201.08239 v3 pith:IATM46SH submitted 2022-01-20 cs.CL cs.AI

Romal Thoppilan , Daniel De Freitas , Jamie Hall , Noam Shazeer , Apoorv Kulshreshtha , Heng-Tze Cheng , Alicia Jin , Taylor Bos
show 52 more authors
Leslie Baker Yu Du YaGuang Li Hongrae Lee Huaixiu Steven Zheng Amin Ghafouri Marcelo Menegali Yanping Huang Maxim Krikun Dmitry Lepikhin James Qin Dehao Chen Yuanzhong Xu Zhifeng Chen Adam Roberts Maarten Bosma Vincent Zhao Yanqi Zhou Chung-Ching Chang Igor Krivokon Will Rusch Marc Pickett Pranesh Srinivasan Laichee Man Kathleen Meier-Hellstern Meredith Ringel Morris Tulsee Doshi Renelito Delos Santos Toju Duke Johnny Soraker Ben Zevenbergen Vinodkumar Prabhakaran Mark Diaz Ben Hutchinson Kristen Olson Alejandra Molina Erin Hoffman-John Josh Lee Lora Aroyo Ravi Rajakumar Alena Butryna Matthew Lamm Viktoriya Kuzmina Joe Fenton Aaron Cohen Rachel Bernstein Ray Kurzweil Blaise Aguera-Arcas Claire Cui Marian Croak Ed Chi Quoc Le
This is my paper · ORCID
classification cs.CLcs.AI
keywords languagemodelsdialogsystemssafetyfactualgroundingfine-tuningexternalknowledgetransformer
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

LaMDA consists of large transformer language models pre-trained on public dialog and web text for conversational use. Scaling the models alone lifts overall response quality but leaves safety and factuality largely unchanged. The work demonstrates that fine-tuning on a modest set of crowdworker annotations for values such as avoiding harm and bias, combined with an interface allowing the model to query information retrieval, translation, and calculation tools, produces measurable gains on both challenges. These gains are quantified with a safety classifier that filters candidate replies and a groundedness metric that checks whether answers rest on verifiable sources rather than plausible invention. The approach is further tested in education and recommendation settings for helpfulness and role consistency.

What carries the argument

A safety classifier fine-tuned on crowd-annotated dialog data that filters responses for alignment with selected human values, paired with an external-tool interface that lets the model call information retrieval, translation, or calculation systems to ground its outputs.

What would settle it

Run the model on prompts involving values outside the annotated set or facts absent from the retrieval, translation, and calculator tools; if the rate of unsafe or ungrounded replies stays as high as in the base model, the claimed improvements do not hold.

Watch

Extended reading notes

Core claim

LaMDA models achieve stronger safety by routing candidate replies through a classifier trained on annotated examples of human values and achieve stronger factual grounding by consulting external knowledge sources during generation rather than relying solely on internal parameters.

Load-bearing premise

The selected human values for annotation and the three chosen external tools are sufficient to cover safety and factuality needs across open-ended real-world conversations.

Editorial extensions

If this is right

  • Dialog systems can filter outputs for consistency with chosen values before they reach users.
  • Models can produce answers that cite or derive from retrieved sources instead of generating from memory alone.
  • Targeted fine-tuning and tool access can outperform further scaling for safety and factuality.
  • Education and recommendation applications show gains in helpfulness and consistency when these methods are applied.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same fine-tuning and tool-use pattern could be applied to other open-ended generation tasks where alignment and verifiability matter.
  • Expanding the set of external tools might allow the model to handle additional reasoning steps not covered by the current three.
  • Safety metrics built on a limited illustrative value set leave room for later expansion or crowdsourced refinement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 3 minor

Summary. The manuscript introduces LaMDA, a family of Transformer-based language models up to 137B parameters pre-trained on 1.56T words of public dialog data and web text. It claims that scaling improves overall dialog quality but yields limited gains on safety and factual grounding. The authors demonstrate that fine-tuning a classifier on a small set of crowdworker-annotated data to filter unsafe responses, combined with enabling the model to consult external tools (information retrieval, translator, calculator), produces improvements on a safety metric derived from an illustrative set of human values and a groundedness metric. The work also includes qualitative explorations of LaMDA in education and content-recommendation domains.

Significance. If the reported gains hold under broader testing, the paper supplies a practical, scalable recipe for mitigating two persistent limitations of large dialog models. The explicit integration of external knowledge sources rather than sole reliance on parametric memory is a clear methodological contribution that later systems have adopted. The scale of the pre-training corpus and the separation of safety fine-tuning from tool-augmented decoding are additional strengths that provide a concrete baseline for subsequent research.

major comments (2)
  1. [Safety and factual grounding sections] Safety and factual-grounding sections: the safety metric is defined over an illustrative set of human values and the groundedness metric depends on the fixed trio of external tools. Because both the training signal and the evaluation metric are constructed from the same limited annotation set and tool interfaces, the measured improvements may be artifacts of the chosen scope rather than robust advances on the broader challenges of safety and factual grounding. An out-of-distribution test set or independently sourced value specification is needed to substantiate the central claim.
  2. [Results sections] Results sections: the manuscript asserts 'significant improvements' yet supplies no error bars, confidence intervals, or statistical significance tests for the safety and groundedness scores. Without these quantities it is impossible to judge whether the observed deltas exceed what could be obtained by alternative fine-tuning regimes or are reliable across random seeds.
minor comments (3)
  1. [Abstract] Abstract: the phrase 'less improvements' is grammatically imprecise and should be replaced by 'smaller improvements' or 'limited improvements'.
  2. [Methods and figures] Figure captions and tool-integration diagrams: the description of how tool calls are interleaved with generation is terse; a short pseudocode snippet or expanded caption would improve reproducibility.
  3. [Related work] Related-work section: several contemporaneous papers on tool-augmented language models and safety fine-tuning are not cited; adding them would better situate the contribution.

Simulated Author's Rebuttal

2 responses · 1 unresolved

Thank you for the constructive feedback on our LaMDA manuscript. We respond to each major comment below, providing clarifications and indicating where revisions can be made to address the concerns.

read point-by-point responses

  1. Referee: [Safety and factual grounding sections] Safety and factual-grounding sections: the safety metric is defined over an illustrative set of human values and the groundedness metric depends on the fixed trio of external tools. Because both the training signal and the evaluation metric are constructed from the same limited annotation set and tool interfaces, the measured improvements may be artifacts of the chosen scope rather than robust advances on the broader challenges of safety and factual grounding. An out-of-distribution test set or independently sourced value specification is needed to substantiate the central claim.

    Authors: The manuscript explicitly describes the safety values as 'illustrative' and the tools as representative examples of external knowledge sources. The improvements demonstrated are specific to this setup, showing that the fine-tuning and tool-use approach can enhance performance on these metrics. We agree that the claims are scoped to the chosen annotations and tools, and we can revise the text to emphasize the illustrative nature and discuss how the framework generalizes to other value sets or tools. However, conducting new out-of-distribution evaluations would require additional crowdworker annotations and experiments not included in the current work. revision: partial

  2. Referee: [Results sections] Results sections: the manuscript asserts 'significant improvements' yet supplies no error bars, confidence intervals, or statistical significance tests for the safety and groundedness scores. Without these quantities it is impossible to judge whether the observed deltas exceed what could be obtained by alternative fine-tuning regimes or are reliable across random seeds.

    Authors: We acknowledge the absence of statistical measures in the reported results. The evaluations were performed using fixed test sets derived from the annotations, and the improvements are presented as direct comparisons. In a revision, we can include error bars estimated via bootstrap resampling or multiple evaluation runs where applicable, and clarify the evaluation methodology to allow assessment of reliability. This will strengthen the presentation without altering the core findings. revision: yes

standing simulated objections not resolved
  • We cannot provide out-of-distribution test sets or independently sourced value specifications, as this would necessitate new data collection efforts beyond the scope of the presented experiments.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity in derivation chain

full rationale

The paper's improvements in safety and factual grounding are demonstrated via fine-tuning on externally sourced crowdworker annotations for an illustrative set of human values and via consultation of independent external tools (IR system, translator, calculator). The safety metric and groundedness metric are defined against these separate annotations and known sources rather than quantities derived from the model's own outputs or fitted parameters. No load-bearing step reduces by construction to self-defined inputs, fitted subsets renamed as predictions, or self-citation chains; the experimental results remain falsifiable against the external benchmarks.

Assumptions & free parameters 0 free parameters · 0 assumptions · 0 invented entities

Abstract contains no explicit mathematical derivations, free parameters, or invented entities; the work is framed as empirical engineering on top of standard transformer pre-training.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LaMDA: Language Models for Dialog Applications." pith.science (2026). https://pith.science/paper/IATM46SH

@misc{pith2026220108239,
  author       = {Pith},
  title        = {Pith review of: LaMDA: Language Models for Dialog Applications},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IATM46SH}},
  note         = {Machine review of arXiv:2201.08239}
}
read the original abstract

We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency.

Discussion (0). Continue with ORCID to comment.

Forward citations

Showing 60 of 167 Pith papers that cite this

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 706 citations worldwide. See all 167 Pith citations

  1. IGC: Integrating a Gated Calculator into an LLM to Solve Arithmetic Tasks Reliably and Efficiently

    cs.LG 2025-01 conditional novelty 8.0 of 10

    A gated calculator module inserted into a frozen Llama 3.1 8B model enables near-perfect arithmetic on BigBench, including multiplication, in a single forward pass without external tools.

  2. AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents

    cs.CR 2024-06 unverdicted novelty 8.0 of 10

    AgentDojo introduces an extensible evaluation framework populated with realistic agent tasks and security test cases to measure prompt injection robustness in tool-using LLM agents.

  3. MusicLM: Generating Music From Text

    cs.SD 2023-01 conditional novelty 8.0 of 10

    MusicLM produces coherent multi-minute 24 kHz music from text prompts using hierarchical sequence-to-sequence modeling and outperforms prior systems in quality and text adherence.

  4. A Recipe for Long-Context Reasoning in Large Language Models via On-Policy Optimization and Distillation

    cs.CL 2026-05 unverdicted novelty 7.0 of 10

    dGRPO merges outcome-based policy optimization with dense teacher guidance from on-policy distillation, yielding more stable long-context reasoning on the new LongBlocks synthetic dataset.

  5. On the Fitness Landscape in the $NK$ Model

    math.PR 2025-08 unverdicted novelty 7.0 of 10

    For the NK fitness landscape with K/N tending to alpha, exact limits for free energy and maximum fitness are identified, together with the geometry of near-fittest peaks.

  6. Every Image Listens, Every Image Dances: Music-Driven Image Animation

    cs.CV 2025-01 conditional novelty 7.0 of 10

    MuseDance animates a reference image into a music-synchronized dance video conditioned only on the audio track and a text description, and contributes a new 2,904-video dataset.

  7. Interleaved-Modal Chain-of-Thought

    cs.CV 2024-11 conditional novelty 7.0 of 10

    Interleaved-Modal Chain-of-Thought (ICoT) inserts attention-selected image patches into a VLM's reasoning trace, improving reasoning accuracy and interpretability over text-only CoT.

  8. Feedback-Guided Extraction of Knowledge Base from Retrieval-Augmented LLM Applications

    cs.CR 2024-11 conditional novelty 7.0 of 10

    An agent-based black-box attack, CopyBreakRAG, progressively extracts over 70% of chunks from RAG knowledge bases by alternating curiosity-driven exploration with reasoning-based exploitation.

  9. Trustworthiness in Retrieval-Augmented Generation Systems: A Survey

    cs.IR 2024-09 unverdicted novelty 7.0 of 10

    Introduces Trust-RAG Compass framework and TRC Bench benchmark to assess RAG trustworthiness across factuality, robustness, fairness, transparency, accountability, and privacy, with evaluations showing performance gap...

  10. Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models

    cs.CL 2023-05 conditional novelty 7.0 of 10

    Plan-and-Solve prompting improves zero-shot LLM reasoning by first creating an explicit plan then executing subtasks, outperforming simple 'think step by step' prompts across ten datasets.

  11. Fast Inference from Transformers via Speculative Decoding

    cs.LG 2022-11 accept novelty 7.0 of 10

    Speculative decoding accelerates exact sampling from large autoregressive models by 2-3x on T5-XXL by running smaller approximation models in parallel to propose token sequences that the large model then verifies in b...

  12. Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks

    cs.CL 2022-11 unverdicted novelty 7.0 of 10

    PoT prompting improves numerical reasoning by having language models write programs executed by a computer instead of performing calculations in natural language chains of thought, with an average 12% gain over CoT.

  13. A Generalist Agent

    cs.AI 2022-05 accept novelty 7.0 of 10

    Gato is a multi-modal, multi-task, multi-embodiment generalist policy using one transformer network to handle text, vision, games, and robotics tasks.

  14. Flamingo: a Visual Language Model for Few-Shot Learning

    cs.CV 2022-04 unverdicted novelty 7.0 of 10

    Flamingo models reach new state-of-the-art few-shot results on image and video tasks by bridging frozen vision and language models with cross-attention layers trained on interleaved web-scale data.

  15. InCoder: A Generative Model for Code Infilling and Synthesis

    cs.SE 2022-04 unverdicted novelty 7.0 of 10

    InCoder is the first generative model to directly perform zero-shot code infilling via bidirectional context from a masked-then-appended training scheme, matching left-to-right models on synthesis while improving on t...

  16. Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language

    cs.CV 2022-04 unverdicted novelty 7.0 of 10

    Socratic Models compose zero-shot multimodal reasoning by prompting pretrained language and vision models to exchange information and enable new capabilities without finetuning.

  17. NAE: Normalizing AutoEncoder

    cs.LG 2026-08 conditional novelty 6.0 of 10

    A conditional surrogate loss that always picks the gradient estimate aligned with the reconstruction loss improves flow autoencoder training and reaches state-of-the-art generative performance on molecules, tabular da...

  18. SPIEval: Evaluating Large Language Models as Mobile Assistants over Scattered Personal Information

    cs.CL 2026-08 conditional novelty 6.0 of 10

    A new human-curated benchmark for LLM mobile assistants shows models struggle to locate scattered personal information, with the best model at 57.3% accuracy.

  19. Toward a Theory of Value in AI Alignment

    cs.AI 2026-08 conditional novelty 6.0 of 10

    A systematic annotation of 94 AI alignment papers shows the field largely equates human values with measurable preferences, rarely defines values, and is increasingly removing humans from alignment evaluation.

  20. Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA

    cs.LG 2026-08 conditional novelty 6.0 of 10

    The released Macaron-V1-Venti model uses a frozen 744B base plus four per-turn-routed LoRA specialists and reports high internal benchmark scores, but it does not demonstrate cross-generation continual-learning gains.

  21. RISE-RL: Rubric-Informed Selective Exploration for Open-Ended Reinforcement Learning

    cs.AI 2026-08 conditional novelty 6.0 of 10

    RISE-RL uses rubric criteria that are frequently missed to generate and filter guided trajectories, weights tokens by natural-policy support, and optimizes them through a separate auxiliary loss, beating Rubric-RL on ...

  22. Can LLM Agents Stick to the Script? A Benchmark for Long-Horizon Consistency in Interactive Narratives

    cs.CL 2026-08 conditional novelty 6.0 of 10

    A new 100-environment benchmark shows that state-of-the-art LLM narrators frequently contradict established story facts and fail to preserve plot commitments over long interactive sessions.

  23. AsymSpec: Efficient Cloud-Edge Speculative Decoding over Asymmetric Networks

    cs.DC 2026-08 conditional novelty 6.0 of 10

    AsymSpec speeds cloud-edge speculative decoding by uploading only token-probability pairs, moving correction traffic to the downlink, and overlapping independent requests instead of same-request runahead.

  24. How Context Attribution Handles What the Model Already Knows

    cs.CL 2026-07 conditional novelty 6.0 of 10

    Context attribution methods cannot disentangle in-context from in-weight knowledge and assign unfaithful scores under overlap; new metrics and WMDP-Cyber++ quantify the failure.

  25. FastTPS: An Optimized Method for LLM Token Phase for AI accelerators

    cs.LG 2026-07 conditional novelty 6.0 of 10

    FastTPS accelerates LLM token-phase inference via reloading-free static KV-cache management, tiled fused RoPE attention, and interlaced-weight MLP fusion, yielding up to 6× speedup at 93% bandwidth on AMD NPUs.

  26. Security--Fidelity Tradeoffs: The Hidden Cost of Prompt Injection Defense

    cs.CR 2026-06 unverdicted novelty 6.0 of 10

    Prompt injection defenses create a security-fidelity tradeoff with no model or defense achieving both high security and high fidelity on the SecFid benchmark across 1,168 examples.

  27. ToxiREX: A Dataset on Toxic REasoning in ConteXt

    cs.CL 2026-06 unverdicted novelty 6.0 of 10

    ToxiREX is a new dataset of 128k Reddit comments in six languages with hierarchical annotations for implicit toxicity in conversational context based on an existing reasoning schema.

  28. In-Context Model Predictive Generation: Open-Vocabulary Motion Synthesis from Language Models to Physics

    cs.RO 2026-06 unverdicted novelty 6.0 of 10

    ICMPG combines LLM-based candidate generation with MPC-style physical simulation and semantic scoring to produce text-driven human motions that are both plausible and faithful.

  29. SharQ: Bridging Activation Sparsity and FP4 Quantization for LLM Inference

    cs.LG 2026-06 unverdicted novelty 6.0 of 10

    SharQ combines input-adaptive N:M sparsity and FP4 quantization via sparse backbone plus dense residual, recovering 43-63% of the NVFP4-to-FP16 accuracy gap on Llama and Qwen models without calibration or retraining.

  30. Provably Efficient Personalized Multi-Objective Bandits with Proactive Conversational Queries

    cs.LG 2026-06 unverdicted novelty 6.0 of 10

    MO-PQUCB hybrid algorithm integrates proactive conversational queries with bandit feedback via shift-invariant regularization to achieve improved regret bounds in personalized multi-objective bandits.

  31. STaR-Quant: State-Time Consistent Post-Training Quantization for Diffusion Large Language Models

    cs.LG 2026-06 unverdicted novelty 6.0 of 10

    STaR-Quant provides a state-time consistent PTQ framework for DLLMs using SGAT and TAC to improve low-bit weight-activation quantization.

  32. CRAFT: Cost-aware Refinement And Front-aware Tuning of Prompts

    cs.CL 2026-06 unverdicted novelty 6.0 of 10

    CRAFT is a Pareto-front prompt optimizer that allocates scarce LLM validation calls to candidates near the current front using accuracy- and cost-oriented generators plus NSGA-II retention.

  33. Throughput-Optimized Networks at Scale

    cs.NI 2026-05 unverdicted novelty 6.0 of 10

    TONS uses linear optimization and heuristics to synthesize deadlock-free network topologies and routing for datacenter AI training, reporting 2.1x and 1.6x geometric mean speedups over best TPU torus variants for unif...

  34. Guaranteed Jailbreaking Defense via Disrupt-and-Rectify Smoothing

    cs.CR 2026-05 unverdicted novelty 6.0 of 10

    DR-Smoothing introduces a disrupt-then-rectify prompt processing scheme into smoothing defenses, delivering tight theoretical bounds on success probability against both token- and prompt-level jailbreaks.

  35. Trivance: Latency-Optimal AllReduce by Shortcutting Multiport Networks

    cs.DC 2026-02 conditional novelty 6.0 of 10

    Trivance is a bidirectional radix-3 AllReduce algorithm that reaches all nodes in log_3 n steps with one-third the congestion of Bruck, with simulation speedups of 5-30%.

  36. When Should LLMs Be Less Specific? Selective Abstraction for Reliable Long-Form Text Generation

    cs.AI 2026-02 conditional novelty 6.0 of 10

    Atom-wise selective abstraction—replacing low-confidence factual claims with higher-confidence, less specific versions—improves the risk-coverage trade-off in long-form generation by up to 27.73% AURC over claim removal.

  37. Your Self-Play Algorithm is Secretly an Adversarial Imitator: Understanding LLM Self-Play through the Lens of Imitation Learning

    cs.LG 2026-02 conditional novelty 6.0 of 10

    LLM self-play finetuning is equivalent to adversarial imitation learning; the chi-squared regularized variant SPIF bounds rewards and improves stability.

  38. From Zipf's Law to Neural Scaling through Heaps' Law and Hilberg's Hypothesis

    cs.IT 2025-12 conditional novelty 6.0 of 10

    Zipf's law, via differential Heaps and Hilberg laws, forces a power-law lower bound on the excess cross entropy of any entropy-bounded foundation model.

  39. ZapGPT: Free-form Language Prompting for Simulated Cellular Control

    cs.AI 2025-09 conditional novelty 6.0 of 10

    ZapGPT evolves a prompt-to-intervention model against a vision-language evaluator and reports that single-prompt training generalizes to unseen free-form language prompts.

  40. DCMI: A Differential Calibration Membership Inference Attack Against Retrieval-Augmented Generation

    cs.CR 2025-09 conditional novelty 6.0 of 10

    DCMI infers RAG database membership by subtracting the system's yes-probability on a perturbed query from the original query, cancelling the interference of non-member retrieved documents.

  41. Benchmarking the Detection of LLMs-Generated Modern Chinese Poetry

    cs.CL 2025-09 conditional novelty 6.0 of 10

    A new modern Chinese poetry detection benchmark shows most current AI-text detectors are unreliable, particularly when LLMs imitate a human style.

  42. Activation Steering Meets Preference Optimization: Defense Against Jailbreaks in Vision Language Models

    cs.CV 2025-08 reject novelty 6.0 of 10

    A proposed VLM defense, SPO-VLM, combines activation steering with sequence-level preference optimization and claims lower jailbreak ASR and toxicity than ASTRA while retaining visual understanding.

  43. Stable Diffusion Models are Secretly Good at Visual In-Context Learning

    cs.CV 2025-08 conditional novelty 6.0 of 10

    A training-free attention recomputation inside Stable Diffusion self-attention enables visual in-context learning across six vision tasks.

  44. CogDual: Enhancing Dual Cognition of LLMs via Reinforcement Learning with Implicit Rule-Based Rewards

    cs.CL 2025-07 conditional novelty 6.0 of 10

    A role-playing LLM that reasons about the scene and its own state before responding, trained with two semantic rewards, beats stronger baselines on role-play benchmarks.

  45. Can Large Language Models Understand As Well As Apply Patent Regulations to Pass a Hands-On Patent Attorney Test?

    cs.CY 2025-07 conditional novelty 6.0 of 10

    Across EQE pre-exam legal questions, OpenAI o1 reached the highest accuracy (0.82), but no tested LLM reached the 0.90 threshold the authors set for passing, and human patent experts found systematic flaws in the mode...

  46. Reinforcement Learning from Human Feedback with High-Confidence Safety Constraints

    cs.LG 2025-06 conditional novelty 6.0 of 10

    HC-RLHF returns an aligned language model only after a held-out safety test certifies, with probability at least 1-delta, that expected harm (as judged by a learned cost model) is below a chosen threshold.

  47. Prompt Candidates, then Distill: A Teacher-Student Framework for LLM-driven Data Annotation

    cs.LG 2025-06 conditional novelty 6.0 of 10

    Prompting LLMs for candidate labels and distilling them into a small model improves annotation accuracy and noise tolerance over single-label annotation.

  48. LAQuer: Localized Attribution Queries in Content-grounded Generation

    cs.CL 2025-06 conditional novelty 6.0 of 10

    LAQuer defines user-initiated, span-level attribution for grounded generation and shows it can cut the text users must read to verify a claim by about two orders of magnitude, at the cost of lower attribution accuracy.

  49. Disentangled Safety Adapters Enable Efficient Guardrails and Flexible Inference-Time Alignment

    cs.LG 2025-05 unverdicted novelty 6.0 of 10

    Disentangled Safety Adapters decouple safety computations from task-optimized LLMs via lightweight adapters, yielding up to 53% better AUC on safety tasks and dynamic inference-time alignment with reduced performance ...

  50. Bayesian Neural Scaling Law Extrapolation with Prior-Data Fitted Networks

    cs.LG 2025-05 conditional novelty 6.0 of 10

    A Prior-data Fitted Network with a scaling-law-specific prior gives better point and uncertainty predictions for neural scaling law extrapolation than MCMC, BNSL, and LC-PFN baselines.

  51. InFact: Informativeness Alignment for Improved LLM Factuality

    cs.CL 2025-05 conditional novelty 6.0 of 10

    InFACT trains LLMs with hierarchical informativeness rewards plus abstention, improving factual precision on QA benchmarks while largely preserving recall.

  52. DeepDialogue: A Multi-Turn Emotionally-Rich Spoken Dialogue Dataset

    cs.CL 2025-05 conditional novelty 6.0 of 10

    DeepDialogue is a new large-scale text-plus-speech dataset of 40,150 multi-turn LLM dialogues with 20 emotion labels across 41 domains, filtered by human-LLM agreement and evaluated for emotional transfer.

  53. Evaluating Intra-firm LLM Alignment Strategies in Business Contexts

    cs.CY 2025-05 conditional novelty 6.0 of 10

    Firms should intentionally align AI assistants' embedded perspectives using supportive, adversarial, or diverse strategies to protect workplace culture and moral norms.

  54. ACCORD: Autoregressive Constraint-satisfying Generation for COmbinatorial Optimization with Routing and Dynamic attention

    cs.LG 2025-05 conditional novelty 6.0 of 10

    ACCORD rewrites combinatorial optimization solutions as step-by-step state updates that keep constraints visible, and this representation raises solution feasibility and lowers optimality gap versus list-of-list and p...

  55. Foundation Models for Geospatial Reasoning: Assessing Capabilities of Large Language Models in Understanding Geometries and Topological Spatial Relations

    cs.CL 2025-05 conditional novelty 6.0 of 10

    Large language models, especially GPT-4 with few-shot prompts, can classify topological spatial relations between WKT-encoded geometries with roughly 0.6 to 0.66 accuracy, though errors cluster near conceptually simil...

  56. DesignFromX: Empowering Consumer-Driven Design Space Exploration through Feature Composition of Referenced Products

    cs.HC 2025-05 conditional novelty 6.0 of 10

    A GenAI-assisted system that decomposes reference product images into design features and lets consumers compose those features into new product designs, improving engagement and exploration in a 24-user study.

  57. Statistical Modeling and Uncertainty Estimation of LLM Inference Systems

    cs.PF 2025-05 conditional novelty 6.0 of 10

    ALA predicts LLM inference throughput for unobserved workloads by fitting an exponential throughput curve per configuration and using XGBoost to predict curve parameters for new configurations.

  58. Seeing Beyond the Scene: Enhancing Vision-Language Models with Interactional Reasoning

    cs.CV 2025-05 conditional novelty 6.0 of 10

    An interaction-augmented scene graph pipeline with chain-of-thought graph construction and reward-based tuning improves VLM reasoning on several benchmarks.

  59. Codifying Character Logic in Role-Playing

    cs.CL 2025-05 conditional novelty 6.0 of 10

    Representing role-play personas as executable if-then-else logic with semantic condition checks improves consistency and lets 1B-parameter models approach the role-play quality of 8B text-prompted models.

  60. An empathic GPT-based chatbot to talk about mental disorders with Spanish teenagers

    cs.HC 2025-05 conditional novelty 6.0 of 10

    A GPT-3-based chatbot with a vulnerable-teenager persona engaged 44 Spanish teenagers in mental-health conversations, with most users opening up emotionally, though the pilot lacked a control group and clinical measures.

See all 167 Pith citations

Reference graph

Works this paper leans on

120 extracted references · 120 canonical work pages · cited by 167 Pith papers (see all)

  1. [1]

    Skip-thought vectors

    Ryan Kiros, Yukun Zhu, Ruslan R Salakhutdinov, Richard Zemel, Raquel Urtasun, Antonio Torralba, and Sanja Fidler. Skip-thought vectors. In Advances in Neural Information Processing Systems, pages 3294–3302, 2015

    work page 2015

  2. [2]

    Semi-supervised sequence learning

    Andrew M Dai and Quoc V Le. Semi-supervised sequence learning. In Advances in Neural Information Processing Systems, 2015

    work page 2015

  3. [3]

    Deep contextualized word representations

    Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettle- moyer. Deep contextualized word representations. In NAACL, 2018

    work page 2018

  4. [5]

    Improving language understanding by generative pre-training

    Alec Radford, Karthik Narasimhan, Tim Salimans, and Ilya Sutskever. Improving language understanding by generative pre-training. https://blog.openai.com/language-unsupervised, 2018

    work page 2018

  5. [6]

    BERT: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of deep bidirectional transformers for language understanding. In NAACL-HLT, 2019

    work page 2019

  6. [7]

    XLNet: Generalized autoregressive pretraining for language understanding

    Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, and Quoc V Le. XLNet: Generalized autoregressive pretraining for language understanding. In NeurIPS, 2019

    work page 2019

  7. [8]

    Albert: A lite bert for self-supervised learning of language representations

    Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. Albert: A lite bert for self-supervised learning of language representations. In International Conference on Learning Representations, 2020. URL https://openreview.net/forum?id=H1eA7AEtvS

    work page 2020

  8. [9]

    RoBERTa: A Robustly Optimized BERT Pretraining Approach

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. RoBERTa: A robustly optimized BERT pretraining approach.arXiv preprint arXiv:1907.11692, 2019

    work page Pith review arXiv 1907

Show all 120 references
  1. [10]

    Le, and Christopher D

    Kevin Clark, Minh-Thang Luong, Quoc V . Le, and Christopher D. Manning. ELECTRA: Pre-training text encoders as discriminators rather than generators. In ICLR, 2020

    2020

  2. [11]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 2020

    2020

  3. [12]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeff...

    2020

  4. [13]

    Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361, 2020

    2001 arXiv

  5. [14]

    Neural responding machine for short-text conversation

    Lifeng Shang, Zhengdong Lu, and Hang Li. Neural responding machine for short-text conversation. In ACL, 2015. 19

    2015

  6. [15]

    A neural network approach to context-sensitive generation of conversational responses

    Alessandro Sordoni, Michel Galley, Michael Auli, Chris Brockett, Yangfeng Ji, Margaret Mitchell, Jian-Yun Nie, Jianfeng Gao, and Bill Dolan. A neural network approach to context-sensitive generation of conversational responses. arXiv preprint arXiv:1506.06714, 2015

    2015

  7. [16]

    Oriol Vinyals and Quoc V . Le. A neural conversational model. In ICML Workshop, 2015

    2015

  8. [17]

    So, Jamie Hall, Noah Fiedel, Romal Thoppilan, Zi Yang, Apoorv Kulshreshtha, Gaurav Nemade, Yifeng Lu, and Quoc V

    Daniel Adiwardana, Minh-Thang Luong, David R. So, Jamie Hall, Noah Fiedel, Romal Thoppilan, Zi Yang, Apoorv Kulshreshtha, Gaurav Nemade, Yifeng Lu, and Quoc V . Le. Towards a human-like open-domain chatbot. arXiv preprint arXiv:2001.09977, 2020

    2001

  9. [18]

    Smith, Y-Lan Boureau, and Jason Weston

    Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, and Jason Weston. Recipes for building an open-domain chatbot. arXiv preprint arXiv:2004.13637, 2020

    2004

  10. [19]

    Recurrent neural network based language model

    Tomas Mikolov, Martin Karafiát, Lukas Burget, Jan Cernock`y, and Sanjeev Khudanpur. Recurrent neural network based language model. In INTERSPEECH, 2010

    2010

  11. [20]

    Generating text with recurrent neural networks

    Ilya Sutskever, James Martens, and Geoffrey E Hinton. Generating text with recurrent neural networks. In ICML, 2011

    2011

  12. [21]

    Exploring the limits of language modeling

    Rafal Józefowicz, Oriol Vinyals, Mike Schuster, Noam Shazeer, and Yonghui Wu. Exploring the limits of language modeling. arXiv preprint arXiv:1602.02410, 2016

    2016

  13. [22]

    Universal language model fine-tuning for text classification

    Jeremy Howard and Sebastian Ruder. Universal language model fine-tuning for text classification. InACL, 2018

    2018

  14. [23]

    Unsupervised representation learning with deep convolutional generative adversarial networks

    Alec Radford, Luke Metz, and Soumith Chintala. Unsupervised representation learning with deep convolutional generative adversarial networks. In ICLR, 2016

    2016

  15. [24]

    Jack W. Rae, Sebastian Borgeaud, Trevor Cai, Katie Millican, Jordan Hoffmann, Francis Song, John Aslanides, Sarah Henderson, Roman Ring, Susannah Young, Eliza Rutherford, Tom Hennigan, Jacob Menick, Albin Cassirer, Richard Powell, George van den Driessche, Lisa Anne Hendricks,...

    2021 arXiv

  16. [25]

    Miller, Kurt Shuster, Jack Urbanek, Douwe Kiela, Arthur Szlam, Iulian Serban, Ryan Lowe, Shrimai Prabhumoye, Alan W

    Emily Dinan, Varvara Logacheva, Valentin Malykh, Alexander H. Miller, Kurt Shuster, Jack Urbanek, Douwe Kiela, Arthur Szlam, Iulian Serban, Ryan Lowe, Shrimai Prabhumoye, Alan W. Black, Alexander I. Rudnicky, Jason Williams, Joelle Pineau, Mikhail S. Burtsev, and Jason Weston....

    2020

  17. [26]

    Personalizing dialogue agents: I have a dog, do you have pets too? ACL, 2018

    Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. Personalizing dialogue agents: I have a dog, do you have pets too? ACL, 2018

    2018

  18. [27]

    A diversity-promoting objective function for neural conversation models

    Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and Bill Dolan. A diversity-promoting objective function for neural conversation models. arXiv preprint arXiv:1510.03055, 2015

    2015

  19. [28]

    Generative deep neural networks for dialogue: A short review

    Iulian Vlad Serban, Ryan Lowe, Laurent Charlin, and Joelle Pineau. Generative deep neural networks for dialogue: A short review. arXiv preprint arXiv:1611.06216, 2016

    2016

  20. [29]

    Transfertransfo: A transfer learning approach for neural network based conversational agents

    Thomas Wolf, Victor Sanh, Julien Chaumond, and Clement Delangue. Transfertransfo: A transfer learning approach for neural network based conversational agents. In NeurIPS Workshop on Conversational AI, 2019

    2019

  21. [30]

    Dialogpt: Large-scale generative pre-training for conversational response generation

    Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, and Bill Dolan. Dialogpt: Large-scale generative pre-training for conversational response generation. arXiv preprint arXiv:1911.00536, 2019

    1911

  22. [31]

    Retrieval augmentation reduces hallucination in conversation

    Kurt Shuster, Spencer Poff, Moya Chen, Douwe Kiela, and Jason Weston. Retrieval augmentation reduces hallucination in conversation. arXiv preprint arXiv:2104.07567, 2021

    2021

  23. [32]

    Adam Roberts, Colin Raffel, and Noam Shazeer. How much knowledge can you pack into the parameters of a language model? In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 5418–5426, November 2020

    2020

  24. [33]

    Nan Du, Yanping Huang, Andrew M. Dai, Simon Tong, Dmitry Lepikhin, Yuanzhong Xu, Maxim Krikun, Yanqi Zhou, Adams Wei Yu, Orhan Firat, Barret Zoph, Liam Fedus, Maarten Bosma, Zongwei Zhou, Tao Wang, Yu Emma Wang, Kellie Webster, Marie Pellat, Kevin Robinson, Kathy Meier-Hellste...

    2021

  25. [34]

    Generalization through memorization: Nearest neighbor language models

    Urvashi Khandelwal, Omer Levy, Dan Jurafsk, Luke Zettlemoyer, and Mike Lewis. Generalization through memorization: Nearest neighbor language models. arXiv preprint arXiv:1911.00172, 2019

    1911

  26. [35]

    Retrieval-augmented generation for knowledge-intensive nlp tasks

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. Retrieval-augmented generation for knowledge-intensive nlp tasks. NeurIPS, 2020

    2020

  27. [36]

    Realm: Retrieval-augmented language model pre-training

    Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Ming-Wei Chang. Realm: Retrieval-augmented language model pre-training. arXiv preprint arXiv:2002.08909, 2020

    2002

  28. [37]

    Leveraging passage retrieval with generative models for open domain question answering

    Gautier Izacard and Edouard Grave. Leveraging passage retrieval with generative models for open domain question answering. arXiv preprint arXiv:2007.01282, 2021

    2007

  29. [38]

    Retrieving and reading: A comprehensive survey on open-domain question answering

    Fengbin Zhu, Wenqiang Lei, Chao Wang, Jianming Zheng, Soujanya Poria, and Tat-Seng Chua. Retrieving and reading: A comprehensive survey on open-domain question answering. arXiv preprint arXiv:2101.00774, 2021

    2021

  30. [39]

    Dense passage retrieval for open-domain question answering

    Vladimir Karpukhin, Barlas O˘guz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen tau Yih. Dense passage retrieval for open-domain question answering. arXiv preprint arXiv:2004.04906, 2020

    2004

  31. [40]

    A modern perspective on query likelihood with deep generative retrieval models

    Oleg Lesota, Navid Rekabsaz, Daniel Cohen, Klaus Antonius Grasserbauer, Carsten Eickhoff, and Markus Schedl. A modern perspective on query likelihood with deep generative retrieval models. arXiv preprint arXiv:2106.13618, 2021

    2021

  32. [41]

    Rae, Erich Elsen, and Laurent Sifre

    Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican, George van den Driessche, Jean-Baptiste Lespiau, Bogdan Damoc, Aidan Clark, Diego de Las Casas, Aurelia Guy, Jacob Menick, Roman Ring, Tom Hennigan, Saffron Huang, Loren Maggiore,...

    2021

  33. [42]

    Tickettalk: Toward human-level performance with end-to-end, transaction-based dialog systems

    Bill Byrne, Karthik Krishnamoorthi, Saravanan Ganesh, and Mihir Sanjay Kale. Tickettalk: Toward human-level performance with end-to-end, transaction-based dialog systems. arXiv preprint arXiv:2012.12458, 2020

    2012

  34. [43]

    Reason first, then respond: Modular generation for knowledge-infused dialogue

    Leonard Adolphs, Kurt Shuster, Jack Urbanek, Arthur Szlam, and Jason Weston. Reason first, then respond: Modular generation for knowledge-infused dialogue. arXiv preprint arXiv:2111.05204, 2021

    2021

  35. [44]

    Webgpt: Browser-assisted question- answering with human feedback

    Reiichiro Nakano, Jacob Hilton, Suchir Balaji, Jeff Wu, Long Ouyang, Christina Kim, Christopher Hesse, Shantanu Jain, Vineet Kosaraju, William Saunders, Xu Jiang, Karl Cobbe, Tyna Eloundou, Gretchen Krueger, Kevin Button, Matthew Knight, Benjamin Chess, and John Schulman. Webg...

    2021 arXiv

  36. [45]

    Internet-augmented dialogue generation

    Mojtaba Komeili, Kurt Shuster, and Jason Weston. Internet-augmented dialogue generation. arXiv preprint arXiv:2107.07566, 2021

    2021

  37. [46]

    Usr: An unsupervised and reference free evaluation metric for dialog generation

    Shikib Mehri and Maxine Eskenazi. Usr: An unsupervised and reference free evaluation metric for dialog generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 681–707, 2020

    2020

  38. [47]

    BLEU: a method for automatic evaluation of machine translation

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei jing Zhu. BLEU: a method for automatic evaluation of machine translation. In ACL, 2002

    2002

  39. [48]

    How NOT to evaluate your dialogue system: An empirical study of unsupervised evaluation metrics for dialogue response generation

    Chia-Wei Liu, Ryan Lowe, Iulian Serban, Mike Noseworthy, Laurent Charlin, and Joelle Pineau. How NOT to evaluate your dialogue system: An empirical study of unsupervised evaluation metrics for dialogue response generation. In Proceedings of the 2016 Conference on Empirical Met...

    2016

  40. [49]

    What makes a good conversation? how controllable attributes affect human judgments

    Abigail See, Stephen Roller, Douwe Kiela, and Jason Weston. What makes a good conversation? how controllable attributes affect human judgments. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language ...

    2019

  41. [50]

    Acute-eval: Improved dialogue evaluation with optimized questions and multi-turn comparisons

    Margaret Li, Jason Weston, and Stephen Roller. Acute-eval: Improved dialogue evaluation with optimized questions and multi-turn comparisons. In NeurIPS workshop on Conversational AI, 2019

    2019

  42. [51]

    Treating dialogue quality evaluation as an anomaly detection problem

    Rostislav Nedelchev, Jens Lehmann, and Ricardo Usbeck. Treating dialogue quality evaluation as an anomaly detection problem. In Proceedings of the 12th Conference on Language Resources and Evaluation , pages 508–512, 2020

    2020

  43. [52]

    On evaluating and comparing conversational agents

    Anu Venkatesh, Chandra Khatri, Ashwin Ram, Fenfei Guo, Raefer Gabriel, Ashish Nagar, Rohit Prasad, Ming Cheng, Behnam Hedayatnia, Angeliki Metallinou, Rahul Goel, Shaohua Yang, and Anirudh Raju. On evaluating and comparing conversational agents. NeurIPS, 2017. 21

    2017

  44. [53]

    Stevie Bergman, Shannon Spruit, Dirk Hovy, Y-Lan Boureau, and Verena Rieser

    Emily Dinan, Gavin Abercrombie, A. Stevie Bergman, Shannon Spruit, Dirk Hovy, Y-Lan Boureau, and Verena Rieser. Anticipating safety issues in e2e conversational ai: Framework and tooling. arXiv preprint arXiv:2107.03451, 2021

    2021

  45. [54]

    Ethical and social risks of harm from language models

    Laura Weidinger, John Mellor, Maribeth Rauh, Conor Griffin, Jonathan Uesato, Po-Sen Huang, Myra Cheng, Mia Glaese, Borja Balle, Atoosa Kasirzadeh, Zac Kenton, Sasha Brown, Will Hawkins, Tom Stepleton, Courtney Biles, Abeba Birhane, Julia Haas, Laura Rimell, Lisa Anne Hendricks,...

    2021 arXiv

  46. [55]

    Dropout distillation

    Samuel Rota Bulò, Lorenzo Porzi, and Peter Kontschieder. Dropout distillation. In ICLR, 2016

    2016

  47. [56]

    The radicalization risks of GPT-3 and advanced neural language models

    Kris McGuffie and Alex Newhouse. The radicalization risks of GPT-3 and advanced neural language models. arXiv preprint arXiv:2009.06807, 2020

    2009

  48. [57]

    Persistent anti-muslim bias in large language models

    Abubakar Abid, Maheen Farooqi, and James Zou. Persistent anti-muslim bias in large language models. arXiv preprint arXiv:2101.05783, 2021

    2021

  49. [58]

    Man is to computer programmer as woman is to homemaker? debiasing word embeddings

    Tolga Bolukbasi, Kai-Wei Chang, James Zou, Venkatesh Saligrama, and Adam Kalai. Man is to computer programmer as woman is to homemaker? debiasing word embeddings. In Advances in Neural Information Processing Systems, 2016

    2016

  50. [59]

    Costa-jussà, and Noe Casas

    Christine Basta, Marta R. Costa-jussà, and Noe Casas. Evaluating the underlying gender bias in contextualized word embeddings. In Proceedings of the First Workshop on Gender Bias in Natural Language Processing , August 2019

    2019

  51. [60]

    Measuring bias in contextualized word representations

    Keita Kurita, Nidhi Vyas, Ayush Pareek, Alan W Black, and Yulia Tsvetkov. Measuring bias in contextualized word representations. In Proceedings of the First Workshop on Gender Bias in Natural Language Processing, August 2019

    2019

  52. [61]

    Lu, Mohamed Abdalla, Matthew McDermott, and Marzyeh Ghassemi

    Haoran Zhang, Amy X. Lu, Mohamed Abdalla, Matthew McDermott, and Marzyeh Ghassemi. Hurtful words: Quantifying biases in clinical contextual word embeddings. In Proceedings of the ACM Conference on Health, Inference, and Learning, 2020

    2020

  53. [62]

    The woman worked as a babysitter: On biases in language generation

    Emily Sheng, Kai-Wei Chang, Premkumar Natarajan, and Nanyun Peng. The woman worked as a babysitter: On biases in language generation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural...

    2019

  54. [63]

    Gender bias in contextualized word embeddings

    Jieyu Zhao, Tianlu Wang, Mark Yatskar, Ryan Cotterell, Vicente Ordonez, and Kai-Wei Chang. Gender bias in contextualized word embeddings. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Techno...

    2019

  55. [64]

    Detecting emergent intersectional biases: Contextualized word embeddings contain a distribution of human-like biases

    Wei Guo and Aylin Caliskan. Detecting emergent intersectional biases: Contextualized word embeddings contain a distribution of human-like biases. arXiv preprint arXiv:2006.03955, 2020

    2006

  56. [65]

    Perturbation sensitivity analysis to detect unintended model biases

    Vinodkumar Prabhakaran, Ben Hutchinson, and Margaret Mitchell. Perturbation sensitivity analysis to detect unintended model biases. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2019

    2019

  57. [66]

    Bowman, and Rachel Rudinger

    Chandler May, Alex Wang, Shikha Bordia, Samuel R. Bowman, and Rachel Rudinger. On measuring social biases in sentence encoders. arXiv preprint arXiv:1903.10561, 2019

    1903

  58. [67]

    Maarten Sap, Dallas Card, Saadia Gabriel, Yejin Choi, and Noah A. Smith. The risk of racial bias in hate speech detection. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019

    2019

  59. [68]

    Shikha Bordia and Samuel R. Bowman. Identifying and reducing gender bias in word-level language models. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Student Research Workshop, 2019

    2019

  60. [69]

    On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 2021

    Emily M Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 2021

    2021

  61. [70]

    Smith, and Yejin Choi

    Maarten Sap, Saadia Gabriel, Lianhui Qin, Dan Jurafsky, Noah A. Smith, and Yejin Choi. Social bias frames: Reasoning about social and power implications of language. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 2020

    2020

  62. [71]

    Social biases in NLP models as barriers for persons with disabilities

    Ben Hutchinson, Vinodkumar Prabhakaran, Emily Denton, Kellie Webster, Yu Zhong, and Stephen Denuyl. Social biases in NLP models as barriers for persons with disabilities. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 2020

    2020

  63. [72]

    Large language models associate muslims with violence

    Abubakar Abid, Maheen Farooqi, and James Zou. Large language models associate muslims with violence. Nature Machine Intelligence, 2021. 22

    2021

  64. [73]

    Extracting training data from large language models

    Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-V oss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, Alina Oprea, and Colin Raffel. Extracting training data from large language models. arXiv preprint arXiv:2012.07805, 2020

    2012

  65. [74]

    Chi, and Alex Beutel

    Sahaj Garg, Vincent Perot, Nicole Limtiaco, Ankur Taly, Ed H. Chi, and Alex Beutel. Counterfactual fairness in text classification through robustness. In Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society, 2019. ISBN 9781450363242

    2019

  66. [75]

    Reducing sentiment bias in language models via counterfactual evaluation

    Po-Sen Huang, Huan Zhang, Ray Jiang, Robert Stanforth, Johannes Welbl, Jack Rae, Vishal Maini, Dani Yogatama, and Pushmeet Kohli. Reducing sentiment bias in language models via counterfactual evaluation. In EMNLP (Findings), 2020

    2020

  67. [76]

    A scalable approach to reducing gender bias in google translate

    Melvin Johnson. A scalable approach to reducing gender bias in google translate. https://ai.googleblog. com/2020/04/a-scalable-approach-to-reducing-gender.html , 2020

    2020

  68. [77]

    Reducing gender bias in word-level language models with a gender-equalizing loss function

    Yusu Qian, Urwa Muaz, Ben Zhang, and Jae Won Hyun. Reducing gender bias in word-level language models with a gender-equalizing loss function. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop, July 2019

    2019

  69. [78]

    Towards debiasing sentence representations

    Paul Pu Liang, Irene Mengze Li, Emily Zheng, Yao Chong Lim, Ruslan Salakhutdinov, and Louis-Philippe Morency. Towards debiasing sentence representations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, July 2020

    2020

  70. [79]

    Recipes for safety in open-domain chatbots

    Margaret Li Y-Lan Boureau Jason Weston Emily Dinan Jing Xu, Da Ju. Recipes for safety in open-domain chatbots. arXiv preprint arXiv:2010.07079, 2020

    2010

  71. [80]

    Smith, and Yejin Choi

    Alisa Liu, Maarten Sap, Ximing Lu, Swabha Swayamdipta, Chandra Bhagavatula, Noah A. Smith, and Yejin Choi. On-the-fly controlled text generation with experts and anti-experts. arXiv preprint arXiv:2105.03023, 2021

    2021

  72. [81]

    Bot-adversarial dialogue for safe conversational agents

    Jing Xu, Da Ju, Margaret Li, Y-Lan Boureau, Jason Weston, and Emily Dinan. Bot-adversarial dialogue for safe conversational agents. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021

    2021

  73. [82]

    Towards understanding and mitigating social biases in language models

    Paul Pu Liang, Chiyu Wu, Louis-Philippe Morency, and Ruslan Salakhutdinov. Towards understanding and mitigating social biases in language models. In ICML, 2021

    2021

  74. [83]

    Ethical challenges in data-driven dialogue systems

    Peter Henderson, Koustuv Sinha, Nicolas Angelard-Gontier, Nan Rosemary Ke, Genevieve Fried, Ryan Lowe, and Joelle Pineau. Ethical challenges in data-driven dialogue systems. In Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society, pages 123–129, 2018

    2018

  75. [84]

    Queens are powerful too: Mitigating gender bias in dialogue generation

    Emily Dinan, Angela Fan, Adina Williams, Jack Urbanek, Douwe Kiela, and Jason Weston. Queens are powerful too: Mitigating gender bias in dialogue generation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020

    2020

  76. [85]

    Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A. Smith. RealToxicityPrompts: Evaluating neural toxic degeneration in language models. In EMNLP (Findings), 2020

    2020

  77. [86]

    Does gender matter? towards fairness in dialogue systems

    Haochen Liu, Jamell Dacon, Wenqi Fan, Hui Liu, Zitao Liu, and Jiliang Tang. Does gender matter? towards fairness in dialogue systems. COLING, 2019

    2019

  78. [87]

    Process for adapting language models to society (PALMS) with values- targeted datasets

    Irene Solaiman and Christy Dennison. Process for adapting language models to society (PALMS) with values- targeted datasets. https://cdn.openai.com/palms.pdf, 2021

    2021

  79. [88]

    Measuring attribution in natural language generation models.arXiv preprint arXiv:2112.12870, 2021

    Hannah Rashkin, Vitaly Nikolaev, Matthew Lamm, Michael Collins, Dipanjan Das, Slav Petrov, Gaurav Singh Tomar, Iulia Turc, and David Reitter. Measuring attribution in natural language generation models.arXiv preprint arXiv:2112.12870, 2021

    2021

  80. [89]

    q2: Evaluating factual consistency in knowledge-grounded dialogues via question generation and question answering

    Or Honovich, Leshem Choshen, Roee Aharoni, Ella Neeman, Idan Szpektor, and Omri Abend. q2: Evaluating factual consistency in knowledge-grounded dialogues via question generation and question answering. arXiv preprint arXiv:2104.08202, 2021

    2021

  81. [90]

    SentencePiece: A simple and language independent subword tokenizer and detokenizer for neural text processing

    Taku Kudo and John Richardson. SentencePiece: A simple and language independent subword tokenizer and detokenizer for neural text processing. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, 2018

    2018

  82. [91]

    Neural machine translation of rare words with subword units

    Rico Sennrich, Barry Haddow, and Alexandra Birch. Neural machine translation of rare words with subword units. arXiv preprint arXiv:1508.07909, 2015

    2015

  83. [92]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. In NeurIPS, 2017

    2017

  84. [93]

    Glu variants improve transformer

    Noam Shazeer. Glu variants improve transformer. arXiv preprint arXiv:2002.05202, 2020

    2002 arXiv

  85. [94]

    Lingvo: a modular and scalable framework for sequence-to-sequence modeling

    Jonathan Shen, Patrick Nguyen, Yonghui Wu, Zhifeng Chen, Mia X Chen, Ye Jia, Anjuli Kannan, Tara Sainath, Yuan Cao, Chung-Cheng Chiu, et al. Lingvo: a modular and scalable framework for sequence-to-sequence modeling. arXiv preprint arXiv:1902.08295, 2019. 23

    1902

  86. [95]

    Hechtman, Yanping Huang, Rahul Joshi, M

    Yuanzhong Xu, HyoukJoong Lee, Dehao Chen, Blake A. Hechtman, Yanping Huang, Rahul Joshi, M. Krikun, Dmitry Lepikhin, Andy Ly, Marcello Maggioni, Ruoming Pang, Noam M. Shazeer, Shibo Wang, Tao Wang, Yonghui Wu, and Zhifeng Chen. Gspmd: General and scalable parallelization for m...

    2021

  87. [96]

    Emily Dinan, Stephen Roller, Kurt Shuster, Angela Fan, Michael Auli, and J. Weston. Wizard of wikipedia: Knowledge-powered conversational agents. arXiv preprint arXiv:1811.01241, 2019

    2019

  88. [97]

    Pitfalls of static language modelling

    Angeliki Lazaridou, Adhiguna Kuncoro, Elena Gribovskaya, Devang Agrawal, Adam Liska, Tayfun Terzi, Mai Gimenez, Cyprien de Masson d’Autume, Sebastian Ruder, Dani Yogatama, Kris Cao, Tomas Kocisky, Susannah Young, and Phil Blunsom. Pitfalls of static language modelling. arXiv p...

    2021

  89. [98]

    Dynamic language models for continuously evolving content

    Spurthi Amba Hombaiah, Tao Chen, Mingyang Zhang, Michael Bendersky, and Marc Najork. Dynamic language models for continuously evolving content. arXiv preprint arXiv:2106.06297, 2021

    2021

  90. [99]

    Truth is a lie: Crowd truth and the seven myths of human annotation

    Lora Aroyo and Chris Welty. Truth is a lie: Crowd truth and the seven myths of human annotation. AI Magazine, 36(1):15–24, Mar. 2015. doi: 10.1609/aimag.v36i1.2564. URL https://ojs.aaai.org/index. php/aimagazine/article/view/2564

    2015 doi

  91. [100]

    Elisa Celis

    Yi Chern Tan and L. Elisa Celis. Assessing social and intersectional biases in contextualized word representations. arXiv preprint arXiv:1911.01485, 2019

    1911

  92. [101]

    Re-imagining algorithmic fairness in india and beyond

    Nithya Sambasivan, Erin Arnesen, Ben Hutchinson, Tulsee Doshi, and Vinodkumar Prabhakaran. Re-imagining algorithmic fairness in india and beyond. arXiv preprint arXiv:2101.09995, 2021

    2021

  93. [102]

    Adversarial training for large neural language models

    Xiaodong Liu, Hao Cheng, Pengcheng He, Weizhu Chen, Yu Wang, Hoifung Poon, and Jianfeng Gao. Adversarial training for large neural language models. arXiv preprint arXiv:2004.08994, 2020

    2004

  94. [103]

    Computer Power and Human Reason: From Judgment to Calculation

    Joseph Weizenbaum. Computer Power and Human Reason: From Judgment to Calculation. W. H. Freeman & Co., New York, 1976. ISBN 0-7167-0463-3

    1976

  95. [104]

    Alignment of language agents

    Zachary Kenton, Tom Everitt, Laura Weidinger, Iason Gabriel, Vladimir Mikulik, and Geoffrey Irving. Alignment of language agents. arXiv preprint arXiv:2103.14659, 2021

    2021

  96. [105]

    Machines and mindlessness: Social responses to computers

    Clifford Nass and Youngme Moon. Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56:81–103, 03 2000. doi: 10.1111/0022-4537.00153

    doi

  97. [106]

    Are machines gender neutral? gender-stereotypic responses to computers with voices

    Clifford Nass, Youngme Moon, and Nancy Green. Are machines gender neutral? gender-stereotypic responses to computers with voices. Journal of Applied Social Psychology, 27(10):864–876, 1997. doi: https://doi.org/ 10.1111/j.1559-1816.1997.tb00275.x. URL https://onlinelibrary.wil...

    1997 doi

  98. [107]

    Participatory interventions for sexual health promotion for adolescents and young adults on the internet: Systematic review

    Philippe Martin, Lorraine Cousin, Serge Gottot, Aurélie Bourmaud, Elise de La Rochebrochard, and Corinne Alberti. Participatory interventions for sexual health promotion for adolescents and young adults on the internet: Systematic review. Journal of Medical Internet Research, ...

    2020

  99. [108]

    Carbon emissions and large neural network training

    David Patterson, Joseph Gonzalez, Quoc Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David So, Maud Texier, and Jeff Dean. Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350, 2021

    2021 arXiv

  100. [109]

    Levesque

    Hector J. Levesque. The winograd schema challenge. In AAAI Spring Symposium: Logical Formalizations of Commonsense Reasoning, 2011. 24 A Safety objectives and data collection A.1 Safety objectives Our research team, which includes people with a wide variety of disciplinary, cu...

    2011

  101. [110]

    This includes giving advice on or otherwise promoting: • Content that could directly facilitate serious and immediate harm to people or animals

    Avoid unintended results that create risks of harm. This includes giving advice on or otherwise promoting: • Content that could directly facilitate serious and immediate harm to people or animals. This includes, but isn’t limited to, dangerous goods, services or activities, an...

  102. [111]

    This includes, • Content that may incite hatred against an individual or group

    Avoid unjust impacts on people, particularly those related to sensitive characteristics associated with systemic discrimination or marginalization such as age, caste, disability, ethnicity, gender identity and expression, nationality, race, immigration status, religion, sex/ge...

  103. [112]

    utterance

    Avoid propagating or reinforcing misinformation that creates risk of harm, as well as opinions likely to incite strong disagreement. This includes, • Theories that are demonstrably false or outdated, as evidenced by, but not limited to legal documents, expert consensus, or oth...

  104. [113]

    Crowdworkers interact with LaMDA to generate conversations on natural, sensitive and adversarial contexts

  105. [114]

    sensitive

    Crowdworkers annotate all the LaMDA-generated utterances given prior context. A.2.1 Conversation generation For the crowdsourcing of the conversation generation task, human participants interacted with LaMDA to generate three types of conversations: natural, sensitive-topic, a...

  106. [115]

    Is the response completely reasonable in context? • If anything seems off—confusing, illogical, out of context, or factually wrong—then rate it as Does not make sense

    Does the response make sense? • Use your common sense here. Is the response completely reasonable in context? • If anything seems off—confusing, illogical, out of context, or factually wrong—then rate it as Does not make sense. • If in doubt, choose Does not make sense

  107. [116]

    I love tennis

    Is the response specific? • You may be asked to assess whether the response is specific to a given context. • For example: – if A says “I love tennis” and B responds “That’s nice”, then mark it as Not specific. That reply could be used in dozens of different contexts. – but if B ...

  108. [117]

    A” and “B

    Is the response interesting? • You may be asked to rate whether or not the response is interesting. • Choose Interesting if the response would likely catch someone’s attention or arouse curiosity; also use that rating for anything insightful, unexpected, or witty. • If the res...

  109. [118]

    Is the conversation understandable? • If the answer is No, you’ll be asked to choose between some possible reasons: – The conversation is nonsensical – The conversation is in a language other than English

  110. [119]

    Yes". • If the response contain even one factual claim, then choose

    Does B’s response contain any factual claims? • Factual claims can either be accurate or inaccurate. • If the response involves information that cannot be objectively verified, such as claims about self, the user or any publicly unrecognizable person, then it is not considered ...

  111. [120]

    No". • If the response contains a link which cannot be verified/opened, then choose

    Is B’s response factually correct? • Use an information retrieval tool to verify each and every factual claim in the response. • Please only focus on objective information that can be verified by reliable sources such as business hours from an official website. • If the response...

  112. [121]

    The Queen

    The first commercially available hy- brid vehicle is the Toyota Prius Around 1900. The first hybrid car was built in the year 1899 by en- gineer Ferdinand Porsche. https: //www.carsdirect.com/green-cars/ a-brief-history-of-hybrid-cars Is a pickled cucumber just a pickle? - Yes. ...

    1900

Pith tools

Reviewed May 12, 2026 · model on record in the stance chip above.